Exchange and Correlation

Direct interaction, exchange, and correlation are three ways in which the same many-electron problem is organized. They are not three fundamental forces.

• The direct Coulomb contribution is the density–density average of the electron–electron interaction.
• Exchange is the crossed contribution forced by fermionic antisymmetry.
• Correlation, in the conventional quantum-chemistry sense used here, is what remains when the exact interacting state is compared with the best single Slater determinant for the same Hamiltonian.

This distinction matters because each layer has different spin dependence, mathematical structure, and approximation error. Hartree theory contains a direct mean field but no exchange. Hartree–Fock contains direct and exchange effects exactly within its determinant trial space, but it omits correlation beyond one determinant. Kohn–Sham density-functional theory groups several residual contributions into an exchange-correlation functional, without making exchange and correlation the same mechanism.

Required background. Helium gives a two-electron spectrum; atomic Pauli restrictions and Slater determinants supply antisymmetric states. The variational principle fixes the sign of ground-state correlation energy.

The general Hartree–Fock derivation and its atomic implementation develop the orbital equations. Here the focus is the pair probabilities and energy contributions those approximations describe.

One Hamiltonian, Different State Structure

In atomic units, with a fixed point nucleus of charge $Z$, the nonrelativistic electronic Hamiltonian is

$$H ={}\sum_{i=1}^{N} \left( -\frac12\nabla_i^2-\frac{Z}{r_i} \right) +\sum_{i<j}\frac{1}{r_{ij}}.$$

The last term is the only electron–electron potential in this model:

$$V_{ee} = \sum_{i<j}\frac{1}{r_{ij}}.$$

Neither an exchange potential nor a correlation potential has been added to the microscopic Hamiltonian. Those names describe structures that appear when $V_{ee}$ is evaluated in an antisymmetric state and when the allowed state family is enlarged beyond a mean-field reference.

The separation is therefore a separation of state-dependent contributions and approximations, not a decomposition into independently acting microscopic forces.

Direct Coulomb Interaction

Let $\phi_a$ and $\phi_b$ be orthonormal spatial orbitals. Their direct Coulomb integral is

$$J_{ab} ={} \iint \frac{ |\phi_a(\mathbf r_1)|^2 |\phi_b(\mathbf r_2)|^2 }{r_{12}} \times d^3r_1\,d^3r_2.$$

This is the classical electrostatic interaction between the two normalized orbital densities. For the repulsive Coulomb kernel,

$$J_{ab}\ge 0.$$

If $h$ is the one-electron part of the Hamiltonian and

$$h_{aa} = \langle\phi_a|h|\phi_a\rangle,$$

then the distinguishable-particle product

$$\phi_a(\mathbf r_1)\phi_b(\mathbf r_2)$$

has energy

$$E_{\mathrm{product}} = h_{aa}+h_{bb}+J_{ab}.$$

The direct term knows about the spatial charge distributions but not about exchange symmetry. It is present for same-spin and opposite-spin electron pairs.

From pair repulsion to a mean field

For a total density $n(\mathbf r)$, the Hartree potential is

$$v_{\mathrm H}(\mathbf r) = \int \frac{n(\mathbf r')}{|\mathbf r-\mathbf r'|} \,d^3r',$$

and the Hartree energy is

$$E_{\mathrm H}[n] = \frac12 \iint \frac{ n(\mathbf r)n(\mathbf r') }{ |\mathbf r-\mathbf r'| } \,d^3r\,d^3r'.$$

The factor $1/2$ avoids double counting pairs. This density product is an average-field object: it does not equal the exact pair density of an interacting atom.

Written in terms of the total one-electron density, $E_{\mathrm H}$ also contains terms in which an orbital density interacts with itself. A correct many-electron treatment must cancel or avoid that self-interaction. Exact determinant exchange performs the cancellation orbital by orbital.

Exchange from Antisymmetry

The exchange integral for the same two spatial orbitals is

$$K_{ab} ={} \iint \phi_a^*(\mathbf r_1) \phi_b(\mathbf r_1) \frac{1}{r_{12}} \times \phi_b^*(\mathbf r_2) \phi_a(\mathbf r_2) \,d^3r_1\,d^3r_2.$$

In two-particle notation,

$$J_{ab} = \langle ab|r_{12}^{-1}|ab\rangle, \qquad K_{ab} = \langle ab|r_{12}^{-1}|ba\rangle.$$

The labels return to the same orbital in $J_{ab}$ and are crossed in $K_{ab}$. Exchange is therefore sensitive to orbital coherence and overlap, not merely to the two orbital densities.

For the Coulomb kernel, $K_{ab}$ is nonnegative. One way to see this is to define the transition density

$$\rho_{ab}(\mathbf r) = \phi_a^*(\mathbf r)\phi_b(\mathbf r).$$

Then $K_{ab}$ is the Coulomb quadratic form of $\rho_{ab}$. In Fourier space,

$$K_{ab} = 4\pi \int \frac{d^3k}{(2\pi)^3} \frac{ |\widetilde\rho_{ab}(\mathbf k)|^2 }{k^2} \ge 0.$$

Orthogonality gives $\widetilde\rho_{ab}(\mathbf 0)=0$, but it does not make the entire transition density vanish.

Spin-adapted two-electron states

The normalized symmetric and antisymmetric spatial combinations are

$$\Phi_{\pm}(1,2) ={}\frac{1}{\sqrt2} \bigl[ \phi_a(1)\phi_b(2) \pm \phi_b(1)\phi_a(2) \bigr].$$

For a spin-independent Hamiltonian,

$$\langle\Phi_{\pm}|H|\Phi_{\pm}\rangle = h_{aa}+h_{bb}+J_{ab}\pm K_{ab}.$$

Fermionic antisymmetry pairs:

• the symmetric spatial state $\Phi_+$ with the antisymmetric spin singlet;
• the antisymmetric spatial state $\Phi_-$ with the symmetric spin triplet.

Thus, at fixed orbitals,

$$\begin{aligned} E_{\mathrm{singlet}} &=h_{aa}+h_{bb}+J_{ab}+K_{ab},\\ E_{\mathrm{triplet}} &=h_{aa}+h_{bb}+J_{ab}-K_{ab}, \end{aligned}$$

and

$$E_{\mathrm{singlet}} -E_{\mathrm{triplet}} = 2K_{ab}.$$

The lower triplet energy in this model is not caused by an additional attractive force. The symmetric and antisymmetric spatial states sample the same Coulomb operator with different interference terms.

A spin-orbital nuance

For spin-orbitals

$$\begin{aligned} \chi_a(x) &=\phi_a(\mathbf r)\omega_{\sigma_a}(s),\\ \chi_b(x) &=\phi_b(\mathbf r)\omega_{\sigma_b}(s). \end{aligned}$$

the exchange integral contains the spin factor

$$\left| \langle \omega_{\sigma_a} | \omega_{\sigma_b} \rangle \right|^2.$$

It vanishes for orthogonal $\alpha$ and $\beta$ spin functions. A single determinant containing $a\alpha$ and $b\beta$ is generally not an eigenstate of total $S^2$; spin adaptation combines determinants to form the singlet and the $M_S=0$ triplet. There is therefore no contradiction between:

• “Fock exchange vanishes between product spin-orbitals with orthogonal spin functions,” and
• “different-orbital singlet and triplet states can be exchange split.”

They refer to different state representations.

Self-interaction cancellation

For a determinant of occupied spin-orbitals, the electron–electron energy is

$$E_{ee}^{\mathrm{det}} = \frac12 \sum_{i,j\in\mathrm{occ}} \left( J_{ij}-K_{ij} \right).$$

Since

$$J_{ii}=K_{ii},$$

every $i=j$ contribution cancels. An electron does not Coulomb-repel itself in exact Hartree–Fock exchange. Approximate density functionals need not preserve this cancellation exactly.

Exchange Splitting in Atoms

The two-orbital result gives the cleanest exchange-splitting model. For helium’s $1s\,2s$ and $1s\,2p$ configurations, it explains why corresponding triplet terms commonly lie below the singlets before smaller relativistic splittings are resolved.

The fixed-orbital formula is not a precision prediction by itself. Singlet and triplet states can have different optimized radial orbitals and different correlation energies. Their measured separation is more accurately organized as

$$\Delta E_{\mathrm{ST}} ={} 2K_{ab} +\Delta E_{\mathrm{relax}} +\Delta E_{\mathrm{corr}} +\Delta E_{\mathrm{rel}} +\Delta E_{\mathrm{QED}} +\cdots.$$

The relative importance of these terms depends on the atom, configuration, and requested accuracy.

More than two electrons

In a many-electron configuration, electrostatic term energies are expressed through angular coefficients and radial Slater–Condon integrals. Several direct and crossed matrix elements contribute, and configuration interaction can mix states with the same exact quantum numbers. There is generally no single number $K$ that equals an observed multiplet splitting.

Exchange supplies an important tendency behind Hund’s first rule: within a configuration, large total spin often permits greater same-spin spatial avoidance and lowers the electrostatic energy. The rule is not a theorem for arbitrary Hamiltonians or configurations. Radial relaxation, near-degeneracy, spin–orbit coupling, and correlation can alter or reverse simple ordering arguments. Hund’s Rules develops the fixed-orbital derivation, the relaxed-energy caveat, and the other two free-atom ordering rules.

Atomic Term Symbols explains the $^{2S+1}L_J$ labels. Pauli Principle in Atoms derives which equivalent-electron terms survive antisymmetrization.

The Exchange Hole

Exchange has a direct pair-probability signature. For a determinant in a collinear spin basis, define the spin-resolved one-body density matrix

$$\gamma_\sigma(\mathbf r,\mathbf r') = \sum_{i\in\mathrm{occ},\sigma} \phi_i(\mathbf r) \phi_i^*(\mathbf r')$$

and the spin density

$$n_\sigma(\mathbf r) = \gamma_\sigma(\mathbf r,\mathbf r).$$

The ordered spin-resolved pair density is

$$n_{\sigma\sigma'}^{(2)} (\mathbf r,\mathbf r') ={} n_\sigma(\mathbf r) n_{\sigma'}(\mathbf r') -\delta_{\sigma\sigma'} \left| \gamma_\sigma(\mathbf r,\mathbf r') \right|^2.$$

The second term removes same-spin pair probability. At coalescence,

$$n_{\sigma\sigma}^{(2)} (\mathbf r,\mathbf r)=0,$$

as required by antisymmetry.

Conditioned on a spin-$\sigma$ electron at $\mathbf r$, the determinant exchange hole is

$$h_x^\sigma(\mathbf r,\mathbf r') = - \frac{ |\gamma_\sigma(\mathbf r,\mathbf r')|^2 }{ n_\sigma(\mathbf r) }.$$

For an idempotent determinant with integer occupations,

$$\int h_x^\sigma(\mathbf r,\mathbf r') \,d^3r' = -1.$$

The conditional density integrates to $N_\sigma-1$, because the electron on which we condition is not counted among its own neighbors. The exchange hole is a statement about pair statistics; it is not a cavity excavated by a mechanical force.

Opposite-spin pairs have no determinant exchange hole in a collinear spin basis. Coulomb interaction can still create an opposite-spin correlation hole.

Correlation Beyond Mean Field

The word correlation has several legitimate meanings:

1. In probability theory, any failure of a joint distribution to factorize is correlation.
2. In a broad many-body usage, the exchange hole itself is a statistical correlation caused by antisymmetry.
3. In conventional electronic-structure theory, electron correlation excludes Hartree–Fock exchange and means the physics missing from the best single determinant.

This page uses the third meaning unless a broader phrase such as “exchange-correlation hole” is written explicitly.

Correlation energy

For the same Hamiltonian, symmetry constraints, and one-particle space, define

$$E_{\mathrm{corr}} = E_0-E_{\mathrm{HF}},$$

where $E_0$ is the exact ground-state energy and $E_{\mathrm{HF}}$ is the global Hartree–Fock minimum. The variational principle gives

$$E_{\mathrm{corr}}\le 0.$$

This definition requires a specified reference problem. Changing the Hamiltonian, basis, frozen core, relativistic treatment, or symmetry restrictions changes the comparison. A merely converged self-consistent solution may be a local minimum or saddle and should not automatically be used as $E_{\mathrm{HF}}$.

Correlation energy is not itself an observable. It is a method-defining energy difference. Two states can have similar correlation energies while differing substantially in contact densities, transition amplitudes, polarizabilities, or other correlation-sensitive quantities.

Wavefunction structure

A correlated state generally requires structure beyond one determinant:

$$|\Psi\rangle ={} C_0|\Phi_0\rangle +\sum_{ia}C_i^a|\Phi_i^a\rangle +\frac14 \sum_{ijab}C_{ij}^{ab} |\Phi_{ij}^{ab}\rangle +\cdots.$$

Alternatively, the ansatz may depend explicitly on interelectronic distances, use coupled-cluster amplitudes, a tensor network, a Green function, or stochastic many-body sampling. The essential point is not a particular algorithm. It is that the pair response is no longer fixed completely by one occupied projector.

Coulomb cusp

For two opposite-spin electrons, the spherical average of an exact nonrelativistic Coulomb wavefunction obeys the electron–electron cusp condition

$$\left. \frac{1}{\Psi} \frac{\partial\Psi}{\partial r_{12}} \right|_{r_{12}=0} = \frac12$$

in atomic units. A finite expansion of smooth orbital products represents this nonanalytic local slope inefficiently. Explicitly correlated factors such as

$$\Psi \sim \Phi_{\mathrm{ref}} \left( 1+\frac12r_{12}+\cdots \right)$$

build the leading opposite-spin coalescence response directly into the state.

The same-spin wavefunction already vanishes linearly at coalescence because of antisymmetry, so its cusp statement must be formulated after factoring out that node. Exchange and Coulomb correlation therefore constrain short-distance behavior in different ways.

Correlation hole

For the exact interacting state, write the conditional deficit relative to its one-electron density as an exchange-correlation hole,

$$h_{xc} = h_x+h_c.$$

For a fixed-$N$ state,

$$\int h_{xc}(\mathbf r,\mathbf r')\,d^3r' = -1.$$

In the standard Kohn–Sham partition, the reference determinant reproduces the same density and its exchange hole also integrates to $-1$. The resulting correlation hole therefore satisfies

$$\int h_c(\mathbf r,\mathbf r')\,d^3r' = 0.$$

Correlation moves conditional probability rather than removing another electron. It produces a negative region near unfavorable pair configurations and compensating positive regions elsewhere. The detailed hole depends on spin, position, state, and interaction strength.

Dynamical and Static Correlation

The distinction is diagnostic rather than mathematically absolute, but it is useful.

Regime	State structure	Atomic signature	Typical warning sign
Dynamical correlation	many small corrections to a dominant determinant	short-range avoidance, radial and angular response, core–valence polarization	slow convergence of energies or contact properties with excitation rank and angular basis
Static correlation	several determinants with comparable weights	near-degenerate open subshells or competing configurations	no determinant has a dominant weight; symmetry breaking lowers a mean-field energy

Hydrogen Molecule provides the canonical molecular comparison: its equilibrium state has substantial dynamical correlation, whereas stretching the bond drives $g^2$ and $u^2$ configurations toward equal importance and exposes static correlation.

Atomic examples

• Helium ground state: one determinant is qualitatively dominant, but radial and angular correlation improve its energy and coalescence properties. Explicit $r_{12}$ coordinates represent those corrections efficiently. This is primarily dynamical correlation.
• Beryllium ground state: the $2s^2$ and $2p^2$ configurations are close enough that a single closed-shell determinant misses important near-degenerate mixing.
• Open $d$ and $f$ shells: several configurations and angular couplings can be competitive, producing substantial multireference structure.
• Core–valence observables: hyperfine constants, isotope shifts, and contact densities can be sensitive to small correlated changes near the nucleus even when their contribution to the total energy is modest.

Relativistic, recoil, QED, finite-nuclear-size, and correlation effects are separate layers of an atomic prediction. A discrepancy with experiment should not be labeled “correlation” until those other approximations and numerical errors have been checked.

Effective Spin Exchange

An ideal Fermi gas retains same-spin exchange anticorrelation even when the pair interaction is switched off. A Coulomb exchange energy requires the interaction as well. In atoms, effective spin Hamiltonians describe energy differences after eliminating orbital degrees of freedom.

Effective exchange Hamiltonians are reduced descriptions

Operators such as

$$H_{\mathrm{eff}} = J_{\mathrm{spin}}\, \frac{\mathbf S_1\cdot\mathbf S_2}{\hbar^2}$$

can faithfully reproduce low-energy singlet–triplet splittings after orbital degrees of freedom are projected out. Here the spin operators have angular-momentum units and $J_{\mathrm{spin}}$ has energy units. This effective parameter should not be confused with the direct Coulomb integral $J_{ab}$. Sign conventions vary; for the displayed convention, $J_{\mathrm{spin}}>0$ places the singlet below the triplet.

Hartree–Fock optimizes direct and exchange contributions over determinants. Correlation methods enlarge the state or response space; their truncation errors concern the same underlying Coulomb problem.

Connection to Density-Functional Theory

The exact Hohenberg–Kohn universal functional is

$$F[n] = \min_{\Psi\to n} \langle\Psi|T+V_{ee}|\Psi\rangle.$$

Kohn–Sham theory rewrites the ground-state energy as

$$E[n] ={} T_s[n] +\int v_{\mathrm{ext}}(\mathbf r) n(\mathbf r)\,d^3r +E_{\mathrm H}[n] +E_{xc}[n].$$

Here $T_s[n]$ is the minimum kinetic energy among admissible noninteracting states reproducing the density. The simple determinant formulas assume a pure-state Kohn–Sham reference; degenerate or ensemble references require the corresponding ensemble definitions. Formally,

$$E_{xc}[n] ={} \bigl(T[n]-T_s[n]\bigr) +\bigl(V_{ee}[n]-E_{\mathrm H}[n]\bigr).$$

Therefore $E_{xc}$ contains more than a correction to the Coulomb energy. It also contains the difference between the interacting and Kohn–Sham kinetic energies.

One may further write

$$E_{xc}[n] = E_x[n]+E_c[n],$$

after specifying a Kohn–Sham exchange convention. Exact Kohn–Sham exchange has the determinant-like crossed form built from occupied Kohn–Sham orbitals. Correlation contains the remaining interaction and kinetic pieces.

The hyphenated phrase exchange-correlation is thus functional bookkeeping, not a claim that exchange and correlation share one physical origin. Approximate local, semilocal, hybrid, and orbital-dependent functionals model this bookkeeping differently. Their results can contain self-interaction, delocalization, static-correlation, and derivative-discontinuity errors even though a term labeled $E_{xc}$ is present.

Kohn–Sham orbitals are auxiliary variables constrained to reproduce a density. Except for results protected by exact theory, their individual eigenvalues and shapes should not automatically be interpreted as exact removal energies or unique physical electron trajectories.

Contact densities and transition amplitudes may converge more slowly than energies. Correlation energy also differs from entanglement: the former is an energy difference relative to Hartree–Fock, while the latter requires a specified subsystem or mode partition.

Exercises

Exercise 1: Derive the exchange splitting

For orthonormal spatial orbitals $a$ and $b$, evaluate $\langle\Phi_\pm|r_{12}^{-1}|\Phi_\pm\rangle$ and identify the spin state paired with each spatial symmetry.

Solution

Expanding the two bra terms against the two ket terms gives two diagonal contributions and two crossed contributions. The normalization factor $1/2$ then yields

$$\langle\Phi_\pm|r_{12}^{-1}|\Phi_\pm\rangle = J_{ab}\pm K_{ab}.$$

The symmetric spatial state $\Phi_+$ must multiply the antisymmetric spin singlet. The antisymmetric spatial state $\Phi_-$ must multiply a symmetric triplet spin function. Hence

$$E_{\mathrm{singlet}} -E_{\mathrm{triplet}} = 2K_{ab}.$$

This is a fixed-orbital result; state-specific relaxation and correlation can change the observed separation.

Exercise 2: Spin selection in a determinant

Show that the exchange integral between $a\alpha$ and $b\beta$ vanishes for a spin-independent interaction, while the direct integral need not vanish.

Solution

The spin part of the exchange matrix element is

$$\langle\alpha|\beta\rangle \langle\beta|\alpha\rangle =0.$$

The direct matrix element instead contains

$$\langle\alpha|\alpha\rangle \langle\beta|\beta\rangle =1.$$

Thus the spin-orbital exchange integral vanishes and the direct Coulomb term remains. This does not make the electrons distinguishable; the determinant is still antisymmetric.

Exercise 3: One-electron self-interaction

Insert a single occupied spin-orbital into the determinant expression for $E_{ee}$. What result should any exact one-electron theory reproduce?

Solution

With only $i=j=1$,

$$E_{ee}^{\mathrm{det}} = \frac12 \left( J_{11}-K_{11} \right).$$

Because $J_{11}=K_{11}$,

$$E_{ee}^{\mathrm{det}}=0.$$

A one-electron system has no electron–electron interaction energy. Hartree energy alone is positive, so exact exchange must cancel it. Failure of an approximate functional to achieve this is one form of self-interaction error.

Exercise 4: Normalize the exchange hole

Use orthonormal occupied orbitals to show that

$$\int |\gamma_\sigma(\mathbf r,\mathbf r')|^2 \,d^3r' = n_\sigma(\mathbf r).$$

Deduce the exchange-hole sum rule.

Solution

Expand both density matrices:

$$\int |\gamma_\sigma(\mathbf r,\mathbf r')|^2 d^3r' = \sum_{ij} \phi_i(\mathbf r)\phi_j^*(\mathbf r) \times \int \phi_i^*(\mathbf r') \phi_j(\mathbf r') d^3r'.$$

Orthonormality reduces the integral to $\delta_{ij}$, leaving

$$\sum_i|\phi_i(\mathbf r)|^2 = n_\sigma(\mathbf r).$$

Therefore

$$\int h_x^\sigma(\mathbf r,\mathbf r') \,d^3r' = -1.$$

Exercise 5: Sign and scope of correlation energy

Why is $E_{\mathrm{corr}}\le0$ for a ground state? Give two changes of convention that prevent correlation energies from being compared directly.

Solution

The exact ground state is varied over a state space containing all normalized determinants, while Hartree–Fock varies over only the determinant subset. Therefore

$$E_0\le E_{\mathrm{HF}},$$

and $E_{\mathrm{corr}}=E_0-E_{\mathrm{HF}}\le0$.

Direct comparison fails if, for example, one value uses a frozen core and another correlates all electrons, or one uses a finite orbital basis while another refers to the complete-basis nonrelativistic limit. Different Hamiltonians, symmetry restrictions, and relativistic models also define different reference problems.

Exercise 6: Identify dynamical and static correlation

Classify the leading difficulty in each case: the helium ground state, stretched $\mathrm H_2$, and an atom with two nearly degenerate configurations of the same symmetry.

Solution

The helium ground state is dominated by one determinant, while many small corrections reproduce radial, angular, and cusp response: primarily dynamical correlation.

At stretched $\mathrm H_2$, the paired $g^2$ and $u^2$ configurations become comparably important in the bonding/antibonding basis, so the restricted one-determinant reference fails qualitatively: static correlation.

Near-degenerate atomic configurations of the same symmetry likewise require substantial configuration mixing and are a static or multireference problem, usually accompanied by additional dynamical correlation.

Exercise 7: What belongs in the DFT remainder?

Starting from

$$F[n]=T[n]+V_{ee}[n],$$

subtract $T_s[n]+E_{\mathrm H}[n]$. Which two physical differences remain in $E_{xc}$?

Solution

The result is

$$E_{xc}[n] ={} \bigl(T[n]-T_s[n]\bigr) +\bigl(V_{ee}[n]-E_{\mathrm H}[n]\bigr).$$

The first line is the interacting kinetic correction. The second contains exchange and Coulomb correlation beyond the Hartree density product. Thus $E_{xc}$ is not merely an empirical version of the quantum-chemistry correlation energy.

References

• E. U. Condon and G. H. Shortley, The Theory of Atomic Spectra, Cambridge University Press (1935).
• R. D. Cowan, The Theory of Atomic Structure and Spectra, University of California Press (1981).
• V. Fock, “Näherungsmethode zur Lösung des quantenmechanischen Mehrkörperproblems,” Zeitschrift für Physik 61, 126–148 (1930), doi:10.1007/BF01340294.
• T. Helgaker, P. Jørgensen, and J. Olsen, Molecular Electronic-Structure Theory, Wiley (2000), doi:10.1002/9781119019572.
• P. Hohenberg and W. Kohn, “Inhomogeneous Electron Gas,” Physical Review 136, B864–B871 (1964), doi:10.1103/PhysRev.136.B864.
• T. Kato, “On the Eigenfunctions of Many-Particle Systems in Quantum Mechanics,” Communications on Pure and Applied Mathematics 10, 151–177 (1957), doi:10.1002/cpa.3160100201.
• W. Kohn and L. J. Sham, “Self-Consistent Equations Including Exchange and Correlation Effects,” Physical Review 140, A1133–A1138 (1965), doi:10.1103/PhysRev.140.A1133.
• P.-O. Löwdin, “Correlation Problem in Many-Electron Quantum Mechanics I,” Advances in Chemical Physics 2, 207–322 (1959), doi:10.1002/9780470143483.ch7.
• NIST, Atomic Spectroscopy: Atomic States, Shells, and Configurations.
• R. G. Parr and W. Yang, Density-Functional Theory of Atoms and Molecules, Oxford University Press (1989).
• J. P. Perdew and A. Zunger, “Self-Interaction Correction to Density-Functional Approximations for Many-Electron Systems,” Physical Review B 23, 5048–5079 (1981), doi:10.1103/PhysRevB.23.5048.
• L. J. Sham, “Exchange and Correlation in Density-Functional Theory,” Physical Review B 32, 3876–3882 (1985), doi:10.1103/PhysRevB.32.3876.
• J. C. Slater, “Note on Hartree’s Method,” Physical Review 35, 210–211 (1930), doi:10.1103/PhysRev.35.210.2.
• A. Szabo and N. S. Ostlund, Modern Quantum Chemistry: Introduction to Advanced Electronic Structure Theory, Dover (1996).