Electron Localization Function

The Electron Localization Function (ELF) is a dimensionless real-space scalar field introduced by Axel D. Becke and Kenneth E. Edgecombe in 1990 and generalized through kinetic-energy-density topology by Andreas Savin and Bernard Silvi in 1991–1994. Constructed directly from the one-particle Kohn-Sham or Hartree-Fock orbitals of a quantum-mechanical wave function, the ELF quantifies the spatial localization of electron pairs by measuring the local repulsion induced by the Pauli exclusion principle between electrons of parallel spin. Because the ELF is scaled to lie within the bounded interval [0, 1], regions where electrons are strongly paired into localized atomic shells, covalent bonds, or non-bonding lone pairs exhibit small excess Pauli repulsion and yield ELF values approaching 1.0, whereas intershell boundaries and delocalized metallic regions exhibit high Pauli repulsion and lower ELF values.

Unlike arbitrary orbital localization schemes (such as Boys or Pipek-Mezey transformations) that depend on a basis-set representation, the Electron Localization Function is invariant with respect to any unitary transformation of occupied molecular orbital theory eigenfunctions. By applying dynamical-systems gradient-field analysis to the ELF scalar field η(r), chemists partition three-dimensional molecular space into non-overlapping basins of attraction that provide a rigorous quantum-mechanical bridge between the Schrödinger equation, Gilbert N. Lewis's electron-pair model, and the valence-shell electron-pair domains of VSEPR theory.

Conditional Same-Spin Pair Probability and the Taylor Expansion

In their original 1990 derivation, Becke and Edgecombe examined the spherically averaged conditional probability P_cond^σσ(r, s) of finding a second electron of spin σ at a distance s from a reference electron of identical spin σ located at position r. Because the antisymmetry of a fermionic determinant enforces a Fermi hole around each reference electron (P_cond^σσ(r, 0) = 0), expanding P_cond^σσ(r, s) in a Taylor series for small interelectronic separation s yields a leading quadratic term:

P_cond^σσ(r, s) = (1/3) D_σ(r) s² + O(s⁴)

The curvature coefficient D_σ(r) controls the rate at which the conditional probability of encountering a second same-spin electron grows with distance s. Where an electron (or an antiparallel-spin electron pair) is well localized in space and isolated from other electrons of the same spin, the Fermi hole is wide and D_σ(r) approaches zero. Conversely, at the boundary between two adjacent localized electron pairs, same-spin electrons are forced into close proximity, making D_σ(r) large.

Excess Pauli Electronic Energy Density Formulation

In closed-shell systems where α and β spin densities are equal (ρ_α = ρ_β = ½ρ), Savin showed that D(r) can be interpreted without reference to a single-determinant Hartree-Fock wave function as the excess Pauli electronic kinetic energy density—the difference between the positive-definite Kohn-Sham electronic kinetic energy density τ(r) of the interacting fermionic system (arising from electron motion and momentum) and the von Weizsäcker bosonic electronic kinetic energy density τ_W(r) that a hypothetical system of non-interacting bosons in motion with the same total electron density ρ(r) would possess:

D(r) = τ(r) − τ_W(r) = ½ ∑_i n_i |∇ψ_i(r)|² − |∇ρ(r)|² / [8ρ(r)]

To normalize D(r) into a bounded, dimensionless index, Becke and Edgecombe divided D(r) by the uniform-gas electronic kinetic energy density D_h(r) of a homogeneous Thomas-Fermi electron gas of identical local density ρ(r) and electron Fermi velocity:

D_h(r) = (3/10) (3π²)^2/3 ρ(r)^5/3

The Electron Localization Function η(r) (commonly denoted ELF) is then defined via a Lorentzian mapping:

ELF(r) = η(r) = 1 / [1 + (D(r) / D_h(r))²]

By construction, the ratio χ(r) = D(r)/D_h(r) is non-negative everywhere:

• ELF(r) = 1.0: Occurs where D(r) = 0, indicating full Pauli localization (as in any one-electron or singlet two-electron system such as H₂ or He, where τ(r) = τ_W(r) everywhere).
• ELF(r) = 0.5: Occurs where D(r) = D_h(r), corresponding to the Pauli delocalization of a homogeneous Thomas-Fermi free-electron gas of the same local density.
• ELF(r) → 0.0: Occurs in regions of high Pauli repulsion (D(r) ≫ D_h(r)), such as the nodal boundary between concentric atomic shells (e.g., between the K and L shells of neon or argon).

Synapticity of Valence Attractors

Applying the gradient vector field ∇η(r) — analogous to the density topology of the Quantum Theory of Atoms in Molecules — partitions molecular space into disjoint basins of attraction, each surrounding a local maximum (attractor) of the ELF and bounded by zero-flux surfaces satisfying ∇η(r) · n(r) = 0. Silvi and Savin classified ELF basins into two fundamental categories:

1. Core Basins, C(A): Surround nuclei with atomic number Z > 2 and encapsulate the tightly bound inner-shell electrons (such as the 1s² K-shell core of carbon, nitrogen, or oxygen, or the K, L, and M shells of heavier elements).
2. Valence Basins, V(A₁, A₂, ...): Occupy the outer molecular envelope and share a common boundary with one or more atomic core basins. Valence basins are classified by their synapticity—the number of atomic cores with which they share a zero-flux boundary:

• Monosynaptic Basins, V(A): Connected to a single atomic core A; these correspond directly to non-bonding lone pairs in Lewis structures.
• Disynaptic Basins, V(A, B): Connected to two atomic cores A and B; these correspond to localized two-center covalent bond domains (single, double, or triple bonds).
• Polysynaptic Basins, V(A, B, C, ...): Connected to three or more cores; these identify multi-center bonding such as the 3-center 2-electron B–H–B bridge bonds in diborane (B₂H₆).

Integrating the electron density ρ(r) over the volume Ω_i of a basin yields the average basin population N̄(Ω_i) = ∫_Ωi ρ(r) d³r, while the covariance of populations between adjacent basins measures quantum-mechanical electron delocalization and bond polarity.

Visualizing Lone Pairs and Toroidal Electron Domains

One of the most important applications of the Electron Localization Function—extensively investigated by Ronald J. Gillespie and Richard F. W. Bader—is the direct quantum-mechanical visualization of VSEPR theory electron domains without invoking localized hybrid orbitals. In water (H₂O), the η(r) = 0.85 localization domain isosurface displays one compact oxygen core sphere C(O), two protonated bonding domains V(O,H) along the O–H axes, and two large kidney-shaped monosynaptic lone-pair domains V(O) oriented in a distorted tetrahedron. Because the V(O) lone-pair attractors are bound to only a single nuclear attractor, their basins occupy a wider angular volume around the oxygen core than the two-center V(O,H) bonding basins, compressing the H–O–H bond angle to 104.45° in agreement with both VSEPR domain repulsion and Bent's rule.

Even more striking is the ELF topology of trigonal bipyramidal AX_nE_5−n hypervalent molecules governed by the three-center four-electron bond. In xenon difluoride (XeF₂, AX₂E₃), instead of three discrete directional sp³d hybrid lobes, the three equatorial lone pairs coalesce due to axial D_∞h cylindrical symmetry into a single continuous equatorial torus (ring attractor) of high electron localization encircling the central xenon core perpendicular to the F–Xe–F axis, containing 6.52 electrons. Similarly, in sulfur tetrafluoride (SF₄, AX₄E), the single equatorial V(S) lone-pair basin expands laterally in the equatorial plane, pushing the axial S–F bonds back to 173.1° and compressing the equatorial F–S–F angle to 101.6°.

Why is the Electron Localization Function equal to 1.0 everywhere in the helium atom and the H₂ molecule?

In any closed-shell one-electron or singlet two-electron system (such as He or H₂), the spatial wave function is described by a single doubly occupied orbital ψ₁(r) = [ρ(r)/2]^1/2. Substituting this relation into the positive-definite Kohn-Sham electronic kinetic energy density of orbital motion yields τ(r) = |∇ρ(r)|²/[8ρ(r)] = τ_W(r). Consequently, the excess Pauli energy density D(r) = τ(r) − τ_W(r) vanishes at every point in space, giving ELF(r) = 1 / (1 + 0) = 1.0 everywhere.

How does the ELF distinguish between covalent, ionic, and metallic bonding?

Covalent bonds produce a distinct disynaptic valence attractor V(A,B) located along the interatomic axis between cores A and B with high localization (ELF ≈ 0.80–0.98) and a basin population near 2.0 electrons per bond order. In strongly ionic bonds (such as NaCl), no shared V(Na,Cl) basin forms; the valence density is transferred into a spherical shell around the anion core. In metallic crystals, valence electrons are delocalized across interstitial space with flat ELF values near the free-electron-gas benchmark of 0.50.

Why does the total population of the six V(S,F) bonding basins in SF₆ sum to less than 12 electrons?

In sulfur hexafluoride (SF₆), each highly electronegative fluorine atom withdraws substantial charge density from the central sulfur atom into a toroidal monosynaptic lone-pair basin V(F) containing approximately 6.56 electrons. Each disynaptic V(S,F) bonding basin contains only ~1.22 electrons (summing to 7.32 bonding electrons around sulfur), demonstrating that hypervalent main-group fluorides stabilize six coordination bonds through polar covalent three-center four-electron bond interactions rather than octet expansion via sulfur 3d orbitals.

1. Becke, A. D., & Edgecombe, K. E. (1990). "A simple measure of electron localization in atomic and molecular systems." The Journal of Chemical Physics, 92(9), 5397–5403. https://doi.org/10.1063/1.458517
2. Silvi, B., & Savin, A. (1994). "Classification of chemical bonds based on topological analysis of electron localization functions." Nature, 371(6499), 683–686. https://doi.org/10.1038/371683a0
3. Savin, A., Nesper, R., Wengert, S., & Fässler, T. F. (1997). "ELF: The Electron Localization Function." Angewandte Chemie International Edition in English, 36(17), 1808–1832. https://doi.org/10.1002/anie.199718081
4. Gillespie, R. J., & Popelier, P. L. A. (2001). Chemical Bonding and Molecular Geometry: From Lewis to Electron Densities. Oxford University Press. https://doi.org/10.1021/ed079p1201