Density functional theory basics: the Hohenberg-Kohn theorems and the Kohn-Sham equations

Intuition Beginner

Every quantum chemistry method you have seen so far works with the wave function -- a complicated object that depends on the coordinates of every electron. For a molecule with $N$ electrons, the wave function $\Psi ({\mathbf{r}}_1,{\mathbf{r}}_2,\dots ,{\mathbf{r}}_N)$ lives in a $3N$-dimensional space. As $N$ grows, storing and manipulating $\Psi$ becomes prohibitively expensive.

Density functional theory (DFT) takes a different approach. Instead of working with the full wave function, DFT uses the electron density $\rho (\mathbf{r})$ -- a function of just three spatial coordinates. The density tells you how many electrons occupy each point in space, regardless of which electron is where. The Hohenberg-Kohn theorems prove that the ground-state density alone determines all ground-state properties of the system, including the total energy. This is a remarkable result: the full $3N$-dimensional wave function contains no more ground-state information than the 3-dimensional density.

The practical question is how to compute the energy from the density. The Kohn-Sham equations provide the answer. They decompose the interacting electron system into a set of one-electron equations -- orbital equations that look like the Hartree-Fock equations but with an extra term called the exchange-correlation functional. If the exact exchange-correlation functional were known, DFT would give the exact ground-state energy. In practice, approximate functionals are used, and the quality of the approximation determines the accuracy of the calculation.

Visual Beginner

The electron density $\rho (\mathbf{r})$ is a scalar field over three-dimensional space. For an atom, it is spherically symmetric and peaks at the nucleus. For a molecule, it accumulates along bonds and around atoms. The total electron count is recovered by integrating the density over all space (summing up the electron population everywhere), which equals $N$.

The Kohn-Sham decomposition replaces the interacting system with a fictitious system of noninteracting electrons that has the same density. Each Kohn-Sham orbital ${\psi }_i(\mathbf{r})$ satisfies a one-electron equation with an effective potential that includes the nuclear attraction, the classical Coulomb repulsion of the electron cloud, and the exchange-correlation potential that accounts for all many-body effects.

Worked example Beginner

A minimal DFT calculation on the helium atom proceeds as follows. Helium has two electrons and $Z=2$. The Kohn-Sham equations for helium reduce to a single orbital equation because both electrons occupy the same spatial orbital ${\psi }_1(\mathbf{r})$ (with opposite spins):

$${\hat{h}}_{\text{KS}}{\psi }_1(\mathbf{r})={\epsilon }_1{\psi }_1(\mathbf{r}),$$

where the Kohn-Sham operator is

$${\hat{h}}_{\text{KS}}=-\frac{{\hslash }^2}{2m_e}\text{(kinetic energy operator)}-\frac{Z{e}^2}{4\pi {\epsilon }_{0}r}+\text{(Coulomb potential integrated over the density)}+v_{\text{xc}}[\rho ](\mathbf{r}).$$

The first term is kinetic energy (the second-derivative operator from quantum mechanics), the second is nuclear attraction, the third is the Coulomb (Hartree) potential from the electron density, and the fourth is the exchange-correlation potential. The density is built from the orbital: $\rho (\mathbf{r})=2|{\psi }_1(\mathbf{r}){|}^2$ (the factor 2 accounts for two electrons).

Self-consistent cycle. Start with a guess for $\rho (\mathbf{r})$ (e.g., a hydrogenic $1s$ orbital with $Z=2$). Build ${\hat{h}}_{\text{KS}}$, solve the orbital equation for ${\psi }_1$, construct a new density ${\rho }_{\text{new}}=2|{\psi }_1{|}^2$, and repeat until the density stops changing. This cycle is the DFT analogue of the Hartree-Fock SCF procedure. Using the local density approximation (LDA) for $v_{\text{xc}}$, the converged energy for helium is approximately $-77.9$ eV, compared with the exact nonrelativistic energy of $-79.0$ eV and the Hartree-Fock energy of $-77.9$ eV.

Check your understanding Beginner

Formal definition Intermediate+

The first Hohenberg-Kohn theorem (existence). For a system of $N$ electrons in an external potential $v_{\text{ext}}(\mathbf{r})$ with a nondegenerate ground state, the ground-state electron density ${\rho }_0(\mathbf{r})$ uniquely determines $v_{\text{ext}}(\mathbf{r})$, up to an additive constant.

Proof (by contradiction). Suppose two external potentials $v(\mathbf{r})$ and $v^{\prime }(\mathbf{r})$, differing by more than a constant, give the same ground-state density ${\rho }_0(\mathbf{r})$. Let $E_0$ and $E_0^{\prime }$ be the corresponding ground-state energies, and ${\Psi }_0$ and ${\Psi }_0^{\prime }$ the ground-state wave functions. By the variational principle applied to $\hat{H}=\hat{T}+{\hat{V}}_{ee}+\hat{V}$:

$$E_0<⟨{\Psi }_0^{\prime }|\hat{H}|{\Psi }_0^{\prime }⟩=⟨{\Psi }_0^{\prime }|{\hat{H}}^{\prime }|{\Psi }_0^{\prime }⟩+\int {\rho }_0(\mathbf{r})[v(\mathbf{r})-v^{\prime }(\mathbf{r})]d\mathbf{r},$$

so $E_0<E_0^{\prime }+\int {\rho }_0(\mathbf{r})[v(\mathbf{r})-v^{\prime }(\mathbf{r})]d\mathbf{r}$. By the same argument with primed and unprimed quantities interchanged:

$$E_0^{\prime }<E_0+\int {\rho }_0(\mathbf{r})[v^{\prime }(\mathbf{r})-v(\mathbf{r})]d\mathbf{r}.$$

Adding these two inequalities gives $E_0+E_0^{\prime }<E_0^{\prime }+E_0$, a contradiction. Therefore $v$ and $v^{\prime }$ cannot differ by more than a constant. $\square$

The second Hohenberg-Kohn theorem (variational principle). Define the universal functional

$$F_{\text{HK}}[\rho ]=⟨\Psi [\rho ]|\hat{T}+{\hat{V}}_{ee}|\Psi [\rho ]⟩,$$

where $\Psi [\rho ]$ is the ground-state wave function of the unique external potential that yields $\rho$. For any valid trial density $\tilde{\rho }(\mathbf{r})$ with $\int \tilde{\rho }d\mathbf{r}=N$:

$$E[v,{\rho }_0]\le E[v,\tilde{\rho }],$$

where $E[v,\rho ]=F_{\text{HK}}[\rho ]+\int v_{\text{ext}}(\mathbf{r})\rho (\mathbf{r})d\mathbf{r}$. Equality holds if and only if $\tilde{\rho }={\rho }_0$.

The Levy constrained-search formulation

The original Hohenberg-Kohn functional requires the density to be $v$-representable (i.e., the ground-state density of some external potential), which is difficult to verify. Levy (1979) and Lieb (1983) introduced the constrained-search functional:

$$F[\rho ]=\underset{\Psi \to \rho }{\min }⟨\Psi |\hat{T}+{\hat{V}}_{ee}|\Psi ⟩,$$

where the minimum is over all antisymmetric $N$-electron wave functions yielding the density $\rho$. This functional is defined for all $N$-representable densities (a much larger class than $v$-representable densities) and reduces to $F_{\text{HK}}$ for $v$-representable densities.

The Kohn-Sham equations

Kohn and Sham (1965) decomposed the universal functional as

$$F[\rho ]=T_s[\rho ]+J[\rho ]+E_{\text{xc}}[\rho ],$$

where $T_s[\rho ]$ is the kinetic energy of a fictitious noninteracting system with density $\rho$, $J[\rho ]=\frac{1}{2}\iint \rho (\mathbf{r})\rho ({\mathbf{r}}^{\prime })/|\mathbf{r}-{\mathbf{r}}^{\prime }|d\mathbf{r}d{\mathbf{r}}^{\prime }$ is the classical Coulomb (Hartree) energy, and $E_{\text{xc}}[\rho ]$ is the exchange-correlation energy that absorbs everything else: the difference between the true kinetic energy and $T_s$, the exchange energy, and the correlation energy.

The Kohn-Sham equations are obtained by minimising the total energy with respect to the orbitals of the noninteracting system:

$$\left[-\frac{{\hslash }^2}{2m_e}{\nabla }^2+v_{\text{ext}}(\mathbf{r})+v_{\text{H}}(\mathbf{r})+v_{\text{xc}}(\mathbf{r})\right]{\psi }_i(\mathbf{r})={\epsilon }_i{\psi }_i(\mathbf{r}),$$

where $v_{\text{H}}(\mathbf{r})=\delta J/\delta \rho =\int \rho ({\mathbf{r}}^{\prime })/|\mathbf{r}-{\mathbf{r}}^{\prime }|d{\mathbf{r}}^{\prime }$ and $v_{\text{xc}}(\mathbf{r})=\delta E_{\text{xc}}/\delta \rho$. The density is reconstructed as $\rho (\mathbf{r})={\sum }_{i=1}^N|{\psi }_i(\mathbf{r}){|}^2$. The equations are solved self-consistently: guess $\rho$, build the potential, solve for orbitals, reconstruct $\rho$, repeat.

The total energy is

$$E_{\text{KS}}=\sum _{i=1}^N{\epsilon }_i-J[\rho ]+E_{\text{xc}}[\rho ]-\int v_{\text{xc}}(\mathbf{r})\rho (\mathbf{r})d\mathbf{r}.$$

The orbital energy sum double-counts the Hartree and exchange-correlation contributions, so the last three terms correct for this.

Key theorem with proof Intermediate+

Theorem (Kohn-Sham equivalence). If the exchange-correlation functional $E_{\text{xc}}[\rho ]$ is known exactly, then the Kohn-Sham scheme yields the exact ground-state density and energy of the interacting $N$-electron system.

Proof. The total energy functional for the interacting system is

$$E[\rho ]=F[\rho ]+\int v_{\text{ext}}(\mathbf{r})\rho (\mathbf{r})d\mathbf{r},$$

where $F[\rho ]={\min }_{\Psi \to \rho }⟨\Psi |\hat{T}+{\hat{V}}_{ee}|\Psi ⟩$ by the Levy constrained search. Decompose $F[\rho ]=T_s[\rho ]+J[\rho ]+E_{\text{xc}}[\rho ]$, where $T_s[\rho ]={\min }_{\{{\psi }_i\}\to \rho }{\sum }_i⟨{\psi }_i|-{\hslash }^2{\nabla }^2/(2m_e)|{\psi }_i⟩$ is the kinetic energy of the noninteracting system.

By the second Hohenberg-Kohn theorem, the ground-state density minimises $E[\rho ]$. Using the chain rule for functional derivatives with the constraint $\int \rho d\mathbf{r}=N$ (enforced by a Lagrange multiplier $\mu$):

$$\frac{\delta T_s}{\delta \rho (\mathbf{r})}+v_{\text{ext}}(\mathbf{r})+\frac{\delta J}{\delta \rho (\mathbf{r})}+\frac{\delta E_{\text{xc}}}{\delta \rho (\mathbf{r})}=\mu .$$

This Euler-Lagrange equation is the stationarity condition for the interacting system. But it is also the Euler-Lagrange equation for a noninteracting system in the effective potential $v_{\text{KS}}(\mathbf{r})=v_{\text{ext}}(\mathbf{r})+v_{\text{H}}(\mathbf{r})+v_{\text{xc}}(\mathbf{r})$, because a noninteracting system satisfies $\delta T_s/\delta \rho (\mathbf{r})=\mu -v_{\text{KS}}(\mathbf{r})$.

By the first Hohenberg-Kohn theorem applied to the noninteracting system, there exists a unique $v_{\text{KS}}$ that yields the density ${\rho }_0$. The solution of the one-electron equations ${\hat{h}}_{\text{KS}}{\psi }_i={\epsilon }_i{\psi }_i$ with $\rho ={\sum }_i|{\psi }_i{|}^2$ reproduces ${\rho }_0$ exactly, provided $E_{\text{xc}}$ is exact. Since ${\rho }_0$ determines all ground-state properties, the Kohn-Sham scheme is exact in principle. $\square$

The practical limitation is that $E_{\text{xc}}[\rho ]$ is unknown. All approximations in DFT flow from the choice of approximate exchange-correlation functional.

The exchange-correlation hole

A useful way to understand $E_{\text{xc}}$ is through the exchange-correlation hole ${\rho }_{\text{xc}}(\mathbf{r},{\mathbf{r}}^{\prime })$, defined by the relation

$${\rho }_2(\mathbf{r},{\mathbf{r}}^{\prime })=\rho (\mathbf{r})[\rho ({\mathbf{r}}^{\prime })+{\rho }_{\text{xc}}(\mathbf{r},{\mathbf{r}}^{\prime })],$$

where ${\rho }_2$ is the pair density. The XC hole represents the reduction in probability of finding an electron at ${\mathbf{r}}^{\prime }$ given one at $\mathbf{r}$, relative to the uncorrelated case. It satisfies the sum rule $\int {\rho }_{\text{xc}}(\mathbf{r},{\mathbf{r}}^{\prime })d{\mathbf{r}}^{\prime }=-1$: on average, each electron excludes exactly one electron from its vicinity.

The exchange-correlation energy can be written as a coupling-constant integral over the XC hole:

$$E_{\text{xc}}[\rho ]=\frac{1}{2}\iint \frac{\rho (\mathbf{r}){\bar{\rho }}_{\text{xc}}(\mathbf{r},{\mathbf{r}}^{\prime })}{|\mathbf{r}-{\mathbf{r}}^{\prime }|}d\mathbf{r}d{\mathbf{r}}^{\prime },$$

where ${\bar{\rho }}_{\text{xc}}$ is the coupling-constant-averaged XC hole. This expression shows that $E_{\text{xc}}$ depends only on a spherical average of the XC hole, which is why even crude approximations to ${\rho }_{\text{xc}}$ can give reasonable energies.

Exercises Intermediate+

The hierarchy of exchange-correlation functionals Master

The accuracy of DFT is entirely determined by the choice of $E_{\text{xc}}[\rho ]$. The functionals are classified by the information they use:

Local density approximation (LDA). $E_{\text{xc}}^{\text{LDA}}[\rho ]=\int \rho (\mathbf{r}){\epsilon }_{\text{xc}}^{\text{unif}}(\rho (\mathbf{r}))d\mathbf{r}$, where ${\epsilon }_{\text{xc}}^{\text{unif}}(\rho )$ is the exchange-correlation energy per particle of a uniform electron gas of density $\rho$. The exchange part is the Dirac exchange: ${\epsilon }_{\text{x}}^{\text{unif}}=-\frac{3}{4}(3\rho /\pi {)}^{1/3}$. The correlation part is taken from quantum Monte Carlo data (Ceperley-Alder, 1980) parametrised by Vosko-Wilk-Nusair (VWN) or Perdew-Zunger (PZ). LDA is exact for a uniform electron gas and works surprisingly well for many systems, but systematically overbinds molecules by 1--2 eV and underestimates bond lengths by 1--2%.

Generalised gradient approximation (GGA). $E_{\text{xc}}^{\text{GGA}}[\rho ]=\int f(\rho (\mathbf{r}),\nabla \rho (\mathbf{r}))d\mathbf{r}$. The functional depends on both the density and its gradient, capturing inhomogeneity effects. Prominent GGAs include Becke-88 (B88) for exchange and Lee-Yang-Parr (LYP) or Perdew-Wang-91 (PW91) for correlation. The popular BLYP functional combines B88 exchange with LYP correlation. GGAs significantly improve atomisation energies, bond lengths, and vibrational frequencies over LDA.

Meta-GGA. Adds dependence on the kinetic energy density $\tau (\mathbf{r})=\frac{1}{2}{\sum }_i|\nabla {\psi }_i(\mathbf{r}){|}^2$ or the Laplacian ${\nabla }^2\rho$. Examples include the Tao-Perdew-Staroverov-Scuseria (TPSS) functional. Meta-GGAs can distinguish between covalent, metallic, and van der Waals bonding through the kinetic energy density.

Hybrid functionals. Mix a fraction of exact (Hartree-Fock) exchange with DFT exchange. The original hybrid is B3LYP:

$$E_{\text{xc}}^{\text{B3LYP}}=(1-a_0)E_{\text{x}}^{\text{LSDA}}+a_0{E}_{\text{x}}^{\text{HF}}+a_{\text{x}}\Delta E_{\text{x}}^{\text{B88}}+(1-a_{\text{c}})E_{\text{c}}^{\text{VWN}}+a_{\text{c}}{E}_{\text{c}}^{\text{LYP}},$$

with $a_0=0.20$, $a_{\text{x}}=0.72$, $a_{\text{c}}=0.81$. B3LYP became the most widely used functional in computational chemistry because it achieves chemical accuracy ($\sim$1 kcal/mol) for thermochemistry of main-group molecules. Its success is partly accidental: the empirical parameters compensate for errors in the GGA components.

Range-separated hybrids. Split the electron-electron interaction into short-range and long-range parts, using exact exchange at long range and DFT exchange at short range. The $\omega$B97X-D and CAM-B3LYP functionals use this approach, significantly improving treatment of charge-transfer excitations and Rydberg states. Long-range exact exchange cures the incorrect $1/R$ decay of the DFT exchange potential.

Double hybrids. Add a perturbative correlation correction (analogous to MP2) on top of the hybrid functional. The B2PLYP functional combines a hybrid DFT calculation with a second-order perturbation theory correction, achieving near-CCSD accuracy at lower cost.

Jacob's ladder. Perdew and Schmidt (2001) organised the functionals into a hierarchy called Jacob's ladder:

Rung	Information used	Examples
1. Hartree (no XC)	$\rho$ only (Coulomb)	None
2. LDA	$\rho$	SVWN, PZ
3. GGA	$\rho$, $\nabla \rho$	BLYP, PBE, PW91
4. Meta-GGA	$\rho$, $\nabla \rho$, $\tau$	TPSS, M06-L
5. Hybrid	Occupied orbitals	B3LYP, PBE0
6. Double hybrid	Virtual orbitals	B2PLYP

Each rung adds information and computational cost. Moving up the ladder generally improves accuracy, but no systematic convergence to the exact functional is guaranteed.

Comparison with Hartree-Fock Master

Hartree-Fock and Kohn-Sham DFT share the same orbital structure: both solve a set of one-electron equations self-consistently. The key difference lies in the treatment of electron-electron interactions.

Property	Hartree-Fock	Kohn-Sham DFT
Fundamental variable	Wave function (Slater determinant)	Electron density
Exchange	Exact (nonlocal)	Approximate (depends on functional)
Correlation	Absent (in HF itself)	Included approximately in $E_{\text{xc}}$
Computational cost	$O(K^4)$ per SCF iteration	$O(K^3)$ for GGA, $O(K^4)$ for hybrids
Self-interaction	Exact cancellation	Incomplete cancellation (LDA, GGA)
Variational	Yes ($E_{\text{HF}}\ge E_{\text{exact}}$)	Only for the exact functional
Band gaps	Overestimated	Underestimated (LDA, GGA)
Dispersion	Absent	Absent in LDA/GGA without corrections

DFT includes electron correlation through $E_{\text{xc}}$ at no additional cost beyond the HF-level SCF cycle. This is DFT's central advantage: correlation comes essentially "for free." The price is that the functional is approximate, and there is no systematic way to improve it (unlike post-HF methods, where one can move from MP2 to CCSD to CCSD(T)).

Hartree-Fock has exact exchange, which makes it self-interaction-free and gives correct asymptotic behaviour of the potential ($v\sim -1/r$ at large distances). Standard LDA and GGA functionals have incorrect asymptotic decay ($v_{\text{x}}^{\text{LDA}}\sim {\rho }^{1/3}\to 0$ exponentially), leading to poor Rydberg excitation energies and anion energies. Hybrid functionals with exact exchange partially fix this.

For thermochemistry of organic molecules, B3LYP/6-31G(d) typically achieves errors of 2--5 kcal/mol for atomisation energies, compared with 5--10 kcal/mol for HF and 2--3 kcal/mol for MP2. For transition metals, strongly correlated systems, and van der Waals complexes, the accuracy of standard functionals degrades significantly, and specialised functionals or post-HF methods are needed.

Strengths, limitations, and failures of DFT Master

Strengths. DFT's favourable cost-to-accuracy ratio makes it the workhorse of computational chemistry. A B3LYP calculation costs roughly the same as an HF calculation but includes correlation. For equilibrium geometries, vibrational frequencies, and reaction energies of main-group molecules, DFT routinely achieves chemical accuracy. The method is available in every major quantum chemistry package (Gaussian, ORCA, Q-Chem, VASP, Quantum ESPRESSO) and scales to systems with hundreds of atoms.

Limitations.

1. No systematic improvability. Unlike the coupled-cluster hierarchy (CCSD, CCSD(T), CCSDT), there is no guaranteed path from an approximate functional to the exact one. Choosing a functional requires judgement and experience.

2. Self-interaction error. All local and semilocal functionals (LDA, GGA, meta-GGA) suffer from self-interaction error, which causes over-delocalisation of electrons. This manifests as underestimated reaction barriers, incorrect dissociation limits for charge-transfer complexes, and qualitatively wrong descriptions of strongly localised states.

3. Static correlation. For systems with near-degeneracies (bond breaking, diradicals, transition metal complexes), the single-determinant Kohn-Sham reference is inappropriate. Multireference DFT methods exist but are less developed than multireference wave function methods.

4. Van der Waals interactions. Standard functionals do not capture long-range dispersion ($\sim C_6/R^6$). Empirical dispersion corrections (DFT-D3, DFT-D4, VV10) add a pairwise $C_6/R^6$ term and are now standard practice.

Known failures. DFT fails or performs poorly for: band gaps of semiconductors and insulators (underestimated by 30--50% with LDA/GGA); charge-transfer excitations (the excitation energy does not have the correct $1/R$ dependence); Rydberg excitations (underestimated by 1--2 eV); reaction barrier heights (overestimated with pure functionals, improved with hybrids); strongly correlated materials (Mott insulators); and thermochemistry involving transition metals (errors of 10--20 kcal/mol are common).

Connections Master

• Many-electron atoms: Hartree-Fock 14.01.03 develops the Hartree-Fock equations and the SCF procedure. The Kohn-Sham equations have the same mathematical structure as the Roothaan equations, with the Fock operator replaced by the Kohn-Sham operator. The SCF cycle, DIIS convergence acceleration, and basis set machinery are shared between HF and DFT.

• Variational methods 14.04.02 provides the Roothaan equations and basis set formalism used in every practical DFT calculation. The Kohn-Sham equations are solved by expanding each orbital in a Gaussian basis set and iterating to self-consistency, exactly as in molecular HF. The second Hohenberg-Kohn theorem is itself a variational principle.

• Perturbation theory 14.04.03 connects to DFT through double-hybrid functionals and the adiabatic connection formula. The MP2 correction in double hybrids uses the same sum-over-virtual-orbitals expression developed for Moller-Plesset theory. The adiabatic connection links the DFT exchange-correlation energy to a coupling-constant integral over the electron-electron interaction strength.

• Molecular orbital theory 14.05.02 provides the qualitative MO picture (bonding, antibonding, $\sigma$, $\pi$) that Kohn-Sham orbitals approximately reproduce. The Kohn-Sham orbitals have no rigorous physical interpretation as individual electron states, but in practice they closely resemble Hartree-Fock orbitals and are used for qualitative MO analysis.

Historical notes Master

Thomas (1927) and Fermi (1927) independently proposed the first density-based theory of atoms, using the kinetic energy of a uniform electron gas as a functional of the density. The Thomas-Fermi model lacks exchange, correlation, and shell structure, and it predicts no molecular binding (Teller's theorem, 1962). It was a conceptual precursor to DFT but had no practical utility for chemistry.

Dirac (1930) added an exchange term proportional to ${\rho }^{4/3}$ to the Thomas-Fermi model, creating the Thomas-Fermi-Dirac model. This was the first local density approximation for exchange. The model still lacked correlation and shell structure.

Hohenberg and Kohn (1964) proved their two theorems, establishing that the density is a valid fundamental variable for many-electron systems. Their paper "Inhomogeneous Electron Gas" in Physical Review is one of the most cited papers in physics. The proof is surprisingly short -- the reductio ad absurdum argument occupies less than a page.

Kohn and Sham (1965) introduced the orbital decomposition that made DFT practical. By splitting the kinetic energy into a noninteracting part (computable from orbitals) and a small correction (absorbed into $E_{\text{xc}}$), they recovered the accuracy of orbital methods while retaining the density as the fundamental variable. The Kohn-Sham approach transformed DFT from a formal theory into a computational method.

The LDA era (1970s-1980s). The local density approximation, parametrised from quantum Monte Carlo data by Ceperley and Alder (1980), became the standard functional for solid-state physics. LDA gave good results for bulk properties of solids (lattice constants within 1--2%, bulk moduli within 10%) but systematically overbound molecules. Its use in chemistry was limited.

The GGA revolution (1988-1992). Becke's 1988 gradient correction for exchange (B88) and Lee, Yang, and Parr's 1988 correlation functional (LYP) dramatically improved DFT accuracy for molecules. Perdew and Wang's PW91 (1991) and Perdew-Burke-Ernzerhof's PBE (1996) provided nonempirical GGAs. These functionals made DFT competitive with MP2 for many chemical applications.

B3LYP and the hybrid era (1993-present). Becke's 1993 three-parameter hybrid functional, combined with LYP correlation, created B3LYP. Its accuracy for main-group thermochemistry (mean absolute error $\sim$2 kcal/mol for the G2 test set) made it the default choice for computational chemistry. B3LYP remains the most-cited functional, though newer functionals (PBE0, $\omega$B97X-D, M06-2X) outperform it for specific applications.

The density functional theory Walter Kohn shared the 1998 Nobel Prize in Chemistry for his development of DFT. John Pople shared the same prize for his development of computational methods in quantum chemistry (the GAUSSIAN program). The joint award recognised the transformative impact of DFT and practical quantum chemistry on the field.

Bibliography Master

• Hohenberg, P. & Kohn, W., "Inhomogeneous Electron Gas," Phys. Rev. 136 (1964), B864-B871.
• Kohn, W. & Sham, L. J., "Self-Consistent Equations Including Exchange and Correlation Effects," Phys. Rev. 140 (1965), A1133-A1138.
• Levy, M., "Universal Variational Functionals of Electron Densities, First-Order Density Matrices, and Natural Spin-Orbitals and Solution of the v-Representability Problem," Proc. Natl. Acad. Sci. USA 76 (1979), 6062-6065.
• Lieb, E. H., "Density Functionals for Coulomb Systems," Int. J. Quantum Chem. 24 (1983), 243-277.
• Ceperley, D. M. & Alder, B. J., "Ground State of the Electron Gas by a Stochastic Method," Phys. Rev. Lett. 45 (1980), 566-569.
• Becke, A. D., "Density-functional Thermochemistry III. The Role of Exact Exchange," J. Chem. Phys. 98 (1993), 5648-5652.
• Lee, C., Yang, W. & Parr, R. G., "Development of the Colle-Salvetti Correlation-Energy Formula into a Functional of the Electron Density," Phys. Rev. B 37 (1988), 785-789.
• Perdew, J. P., Burke, K. & Ernzerhof, M., "Generalized Gradient Approximation Made Simple," Phys. Rev. Lett. 77 (1996), 3865-3868.
• Parr, R. G. & Yang, W., Density-Functional Theory of Atoms and Molecules (Oxford, 1989).
• Dreizler, R. M. & Gross, E. K. U., Density Functional Theory (Springer, 1990).
• Koch, W. & Holthausen, M. C., A Chemist's Guide to Density Functional Theory, 2e (Wiley-VCH, 2001).
• Levine, I. N., Quantum Chemistry, 7e (Pearson, 2014), Ch. 16.
• Burke, K., "Perspective on Density Functional Theory," J. Chem. Phys. 136 (2012), 150901.