Introduction to
Density Functional Theory
C. David Sherrill
Georgia Institute of Technology
The Big Picture
DFT improves upon Hartree-Fock by including an
approximate treatment of the correlated motions of
electrons (these are treated in Hartree-Fock in only an
averaged sense)
Treatment of electron correlation is much cheaper than in
correlated wavefunction methods like MP2, CCSD, CCSD(T)
Although there is a loose ordering of density functionals
from less sophisticated to more sophisticated (e.g., LDA
to GGA to hybrid, etc.), in practice there is no reliable way
to improve your computation by going to the next better
functional. By contrast, this is possible with wavefunction
methods: one almost always has CCSD(T)>CCSD>MP2>HF.
Hohenberg and Kohn
First Hohenberg-Kohn theorem: The ground
state properties of a many-electron system
depend only on the electronic density n(x,y,z)
Second Hohenberg-Kohn theorem: The
correct ground state density for a system is
the one that minimizes the total energy
through the functional E[n(x,y,z)]
A functional is just a function that depends on
a function
Form of the Density Functional
So whats the density functional actually look
like?
The Coulomb interaction for a given density
interacting the nuclei is very straightforward to
compute; so is the Coulomb interaction of the
density with itself (J term)
Coulomb (J) terms are great, but we also need to
account for electron antisymmetry (exchange
effects) and electron correlation effects
Additionally, not clear how to compute kinetic
energy as a function of the density
Form of the Density Functional
Kohn and Sham (KS)
Compute the kinetic energy of a density by
assuming that the density corresponds to a
wavefunction consisting of a single Slater
determinant (non-interacting limit): we know
how to compute the kinetic energy of a Slater
determinant (orbitals) --- looks same as HartreeFock Theory
This procedure is called Kohn-Sham DFT and is
the most common approach (although doesnt
work well for extremely large systems due to
computational cost)
Kohn-Sham Kinetic Energy
Exchange-Correlation Functional
We can compute every piece of a Kohn-Sham DFT
energy exactly except for the exchangecorrelation piece, Exc[].
Unfortunately the exact exchange-correlation
energy functional is not known and is probably so
complicated that even if it were known it would
not be computationally useful
Hence, use various approximate exchangecorrelation functionals (S-VWN, B3LYP, etc.)
Kohn-Sham DFT Self-Consistent-Field
Equations
Much like Hartree-Fock but theres an extra exchangecorrelation piece that adds approximate electron
correlation and can handle the exchange term differently
Observations on KS DFT
Cost is similar to HF (similar equations) but
quality can be better because correlation is
built in through the correlation functional
Cost can actually be cheaper than HF if we
replace the expensive, long-range exchange
integrals (K terms) from HF with a shorterrange exchange potential (which however
might not be as accurate)
The Exchange-Correlation Potential
Often one breaks up Exc into exchange and correlation
parts and writes the in terms of the energy per particle, x
and c
Hierarchy of DFT Exchange-Correlation
Functionals
Local density approximation (LDA): Functional depends
only on the (local) density at a given point. Example: SVWN
Gradient-corrected approximation (GGA): Functional
depends on local density and its gradient. Examples:
PW91 and LYP correlation functionals, B88 exchange
functional
Meta-GGA: Functional depends on density, its gradient,
and its second derivative. Example: M06-L
Hybrid DFT: Mixes in Hartree-Fock exchange. Most
popular example: B3LYP (hybrid GGA). M05-2X and
M06-2X are hybrid meta-GGAs.
Local (Spin) Density Approximation
Based on results for the uniform electron gas
as a model, Dirac introduced this form for the
exchange functional (called Slater exchange,
abbreviated S):
Local (Spin) Density Approximation
If the and densities differ (e.g., open-shell
systems), use the local spin density
approximation (LSDA)
Local (Spin) Density Approximation
For the correlation energy, again use the uniform
electron gas as a model. Its correlation energy
was determined numerically by Monte Carlo
simulations and fit to an analytic form by Vosko,
Wilk, and Nusair (VWN), to give cVWN. L(S)DA
usually implies VWN correlation
More technical name for L(S)DA is S-VWN (Slater
exchange plus Vosko, Wilk, Nusair correlation)
Electron correlation can be overestimated by a
factor of 2 when using VWN. Bond strengths too
large. Need better approximations
Generalized Gradient Approximations
(GGAs)
The uniform electron gas isnt such a great model
because in molecules, the electron density can vary
rapidly over a small region of space
One way to improve over LSDA is to make the
functional depend on both the density and the
gradient of the density
This leads to gradient corrected functionals or
generalized gradient approximation (GGA)
functionals
Also sometimes called nonlocal functionals but this is
a bad name that is fortunately falling out of favor now:
the density and its gradient still supply only local
information
Examples of GGAs
Exchange: PW86, B88 (B), PW91
Correlation: LYP
GGAs improve over L(S)DA but are still not necessarily
very accurate; one reason is the exchange potential
doesnt necessarily have the correct qualitative
behavior
B88 has the correct -1/r asymptotic behavior of the
energy density but not the overall exchange potential
Hybrid Funtionals
Introduced by Axel Becke in 1993 to provide improved
results
Inspired by the Adiabatic Connection Formula, a formally
exact way to get Exc; interpolate between the real
(interacting) system and a formal noninteracting limit
where a single Slater determinant made of KS orbitals
exactly solves the Schrdinger equation
For 0 < < 1, Vext() is adjusted to that the same density is
obtained for =0 and =1
Adiabatic Connection Formula
If Vxc is linear in , can approximate as
=0: only exchange, no correlation; 0 is a
Slater determinant, can use HF equation for
exchange energy
=1: Real system, unknown solution. Can
approximate with LSDA and add to =0
part to give half and half
Beckes 3-Parameter Hybrids
Since GGAs are better than LSDA, could also use
a functional form like this (B3):
B3LYP uses this form, with LYP for EcGGA
B3LYP introduced in 1994 by Stephens, Devlin,
Chabalowski, and Frisch; their computer code did
not match their paper (confusion about VWN-1
vs VWN-5 correlation); led to problems deciding
about what is the real B3LYP
What you need to specify to run a DFT
computation
Molecule
Molecular charge
Spin multiplicity
Basis set
Exchange functional: S, B, B3, etc.
Correlation functional: LYP, PW91, etc.
Observations
DFT good for geometries, often not as good for energies
B3LYP works really well and is hard to beat (usually)
Minnesota functionals (M05-2X, M06-2X, M06-L, etc) seem to work
well also (but can be sensitive to the numerical integration grid)
Barrier heights often underestimated
Totally fails for non-covalent interactions (just like Hartree-Fock); fix
with DFT-D or XDM or vdW-DFT
Can have large errors for excitation energies to Rydberg excited
states (fix with asymptotically-corrected functionals like CAMB3LYP) (DFT for excited states is called time-dependent DFT, or
TDDFT)
Can totally fail for charge-transfer states (fix with range-separated
hybrids that include HF exchange at long range only)
Can get wrong energetic ordering of spin states of metals (just like
Hartree-Fock)
