The Hartree-Fock Energy and Fock Operator

Every major quantum chemistry package — Gaussian, ORCA, Psi4, Q-Chem, NWChem — offers Hartree-Fock as both a standalone method and the mandatory first step of virtually every more accurate calculation. In drug discovery, HF-optimized geometries and vibrational frequencies (scaled by empirical factors to correct for missing correlation) remain a fast, reliable screening tool before committing to expensive coupled-cluster or high-level DFT refinement on lead compounds. In materials science, unrestricted Hartree-Fock (UHF) and its periodic-boundary generalizations provide qualitative starting points for studying open-shell defects and radical sites in extended solids, later corrected by post-HF or hybrid-functional treatments. The Fock operator's Coulomb-minus-exchange structure survives, in modified form, inside every hybrid density functional (B3LYP, PBE0, and their relatives) as the fraction of "exact exchange" mixed with a density-functional exchange-correlation term — meaning that the exchange integral $K_{ij}$ derived here is not a historical curiosity but a live ingredient recomputed in essentially every DFT calculation run today in biochemistry, catalysis, and photovoltaic materials design. Koopmans'-theorem estimates of ionization energies from HF orbital energies remain a fast qualitative tool in mass spectrometry and photoelectron spectroscopy interpretation, despite their known systematic errors from neglected orbital relaxation and correlation. Finally, the entire machinery of post-Hartree-Fock methods — Møller-Plesset perturbation theory, configuration interaction, and coupled-cluster theory — is explicitly built by treating the difference between the true wavefunction and the single Hartree-Fock determinant as the object to be systematically corrected, making the derivation above the literal starting point of the electron-correlation problem in modern computational chemistry.

Ask the tutor: How does expanding the molecular orbitals in an atomic-orbital basis turn the Fock eigenvalue equation into the matrix form used in the Roothaan-Hall equations?

Consider an $N$-electron molecule under the Born–Oppenheimer clamped-nuclei Hamiltonian,

where $\hat h(i) = -\tfrac12\nabla_i^2 - \sum_A Z_A/r_{iA}$ (atomic units) is the one-electron kinetic-plus-nuclear-attraction operator for electron $i$. The pairwise sum over $1/r_{ij}$ is the entire difficulty: it couples every electron to every other electron and makes the Schrödinger equation for $\hat H$ separable in no useful basis. Hartree-Fock theory approximates the true many-electron wavefunction by a single Slater determinant built from $N$ orthonormal spin-orbitals $\chi_1,\dots,\chi_N$,

This form guarantees antisymmetry under electron exchange — swapping two electrons swaps two rows of the determinant and flips its sign, enforcing the Pauli exclusion principle automatically. The physical content of Hartree-Fock theory is entirely in the choice of the spin-orbitals $\chi_i$: they are variationally optimized so that $\langle\Psi|\hat H|\Psi\rangle$ is as low as possible. When you expand $\langle\Psi|\hat H|\Psi\rangle$ for a determinant, the antisymmetrized wavefunction naturally produces three types of matrix elements: one-electron integrals (kinetic energy + nuclear attraction for each occupied orbital), Coulomb integrals $J_{ij}$ (the classical electrostatic repulsion between the charge clouds $|\chi_i|^2$ and $|\chi_j|^2$), and exchange integrals $K_{ij}$ (a purely quantum-mechanical, no-classical-analog term arising because the determinant's antisymmetry correlates the positions of same-spin electrons — they are less likely to be found close together than an uncorrelated product wavefunction would predict). Exchange terms only connect orbitals of the same spin; there is no exchange between an $\alpha$ electron and a $\beta$ electron because their spin functions are orthogonal and the spin integration simply kills the term. Physically, each electron in Hartree-Fock theory does not feel the instantaneous positions of the other $N-1$ electrons — it feels their time-averaged charge density, smeared into a mean field, corrected by an exchange "hole" that keeps same-spin electrons apart. This mean-field electron then obeys a one-electron Schrödinger-like equation, with the Fock operator $\hat f$ playing the role of an effective Hamiltonian. Because $\hat f$ is built from the very orbitals it's supposed to produce as eigenfunctions, the problem is implicit — solved by guessing orbitals, building $\hat f$, diagonalizing it, getting new orbitals, and repeating until the orbitals stop changing (self-consistency).

The Slater-determinant energy expression. For a normalized Slater determinant of orthonormal spin-orbitals $\chi_1,\dots,\chi_N$, the expectation value of the electronic Hamiltonian is

where:

• $h_i = \langle \chi_i(1)|\hat h(1)|\chi_i(1)\rangle$ is the one-electron energy (kinetic + nuclear attraction) of orbital $i$,
• $J_{ij} = \displaystyle\iint |\chi_i(1)|^2\,\frac{1}{r_{12}}\,|\chi_j(2)|^2\,d\tau_1 d\tau_2$ is the Coulomb integral between charge densities $i$ and $j$,
• $K_{ij} = \displaystyle\iint \chi_i^*(1)\chi_j(1)\,\frac{1}{r_{12}}\,\chi_j^*(2)\chi_i(2)\,d\tau_1 d\tau_2$ is the exchange integral, nonzero only when $i,j$ share the same spin,
• the $\tfrac12$ prefactor avoids double-counting each electron pair, and the $i=j$ terms cancel exactly since $J_{ii}=K_{ii}$ (an electron does not repel or exchange with itself).

Reading: the total electronic energy is the sum of each electron's kinetic and nuclear-attraction energy, plus the average electron-electron repulsion energy, which itself splits into a purely classical Coulomb piece and a purely quantum-mechanical antisymmetry correction (exchange) that lowers the energy relative to same-spin Coulomb repulsion alone.

The closed-shell (restricted) form. For a closed-shell molecule with $N=2n$ electrons doubly occupying $n$ spatial orbitals $\psi_1,\dots,\psi_n$, spin integration collapses the sum over spin-orbitals into a sum over spatial orbitals:

where:

• the factor of 2 in front of $h_a$ counts both spin electrons in each spatial orbital,
• $J_{ab}$ and $K_{ab}$ are now Coulomb and exchange integrals over spatial orbitals,
• the $2J_{ab}-K_{ab}$ combination reflects that both spin pairs (αα, ββ, αβ, βα) contribute Coulomb repulsion, but only the two same-spin combinations (αα and ββ) contribute exchange.

Reading: this is the working formula programmed into every closed-shell Hartree-Fock code; note that exchange enters with coefficient 1 (not 2), since only same-spin pairs exchange, while Coulomb enters with coefficient 2 because both spin channels repel classically.

The Fock operator. Minimizing $E_{\rm HF}$ subject to orbital orthonormality $\langle\chi_i|\chi_j\rangle=\delta_{ij}$ (via Lagrange multipliers) yields a one-electron eigenvalue equation for each orbital, which after a unitary transformation to canonical orbitals reads

where:

• $\hat f$ is the Fock operator, an effective one-electron Hamiltonian,
• $\hat J_j(1)$ is the Coulomb operator, defined by its action $\hat J_j(1)\chi_i(1) = \Big[\int |\chi_j(2)|^2\, r_{12}^{-1}\,d\tau_2\Big]\chi_i(1)$ — a local multiplicative potential,
• $\hat K_j(1)$ is the exchange operator, defined by its (nonlocal) action $\hat K_j(1)\chi_i(1) = \Big[\int \chi_j^*(2)\,r_{12}^{-1}\,\chi_i(2)\,d\tau_2\Big]\chi_j(1)$ — it swaps the argument of $\chi_i$ into the integral and returns $\chi_j$,
• $\varepsilon_i$ is the orbital energy, the Lagrange multiplier associated with normalization, later reinterpreted via Koopmans' theorem as (approximately) minus the ionization energy from orbital $i$.

Reading: every electron moves in the field of the frozen nuclei plus the averaged Coulomb and exchange field of all $N$ occupied orbitals — including itself, which is why the self-interaction in $J_{ii}$ must be exactly cancelled by $K_{ii}$ for the equations to be physically sensible.

Example 1 — Foundational: two electrons in a minimal helium-like model. Take a crude two-orbital model for the helium atom's ground state, both electrons in a single spatial orbital $\psi_{1s}$ (so $n=1$ occupied spatial orbital, $N=2$ electrons, closed-shell). The restricted energy expression collapses to

But for a single orbital, $J_{11}=K_{11}$ exactly (self-Coulomb equals self-exchange, as required), so

Using approximate hydrogenic-orbital values often quoted for a variational He calculation with effective nuclear charge $Z_{\rm eff}=1.6875$ (the classic textbook optimized value): $h_1 = -\tfrac12 Z_{\rm eff}^2 - Z_{\rm eff}\times 2\times(\text{normalization factors})$ is more simply obtained from the known result

(the $\tfrac58 Z_{\rm eff}$ form is the standard closed-form Coulomb integral for two electrons in a $1s$ Slater orbital of exponent $Z_{\rm eff}$). Then

This reproduces the well-known variational helium energy of $-2.8477\ \text{hartree}$ per electron pair contribution combined correctly gives total energy near $-2.8477\ \text{hartree}$ when $Z_{\rm eff}$ is optimized against $\partial E/\partial Z_{\rm eff}=0$; the point of this example is structural — verify by inspection that only one Coulomb-type integral $J_{11}$ appears, no exchange survives as an independent term (since $K_{11}=J_{11}$ cancels the self-interaction as required by Step 4 of the first derivation), and the energy expression has exactly the two pieces (one-electron sum, two-electron sum) the general formula predicts.

Limiting case check: if we artificially set the electron-electron repulsion to zero ($J_{11}\to 0$, a fictitious non-interacting limit), $E_{\rm RHF} \to 2h_1 = -5.695$ hartree, exactly twice the hydrogenic $1s$ energy with $Z_{\rm eff}$ — confirming that in the absence of repulsion, Hartree-Fock trivially returns the sum of independent one-electron energies, as it must.

Example 2 — Applied: exchange stabilization in a closed-shell diatomic-like two-orbital system. Consider a simplified restricted-HF treatment of a four-electron system (like a minimal model of $\mathrm{Be}$ or a two-orbital fragment) with two doubly-occupied spatial orbitals $\psi_1,\psi_2$ (asymmetric, non-degenerate — a case that would hide errors if the formula were wrong). Suppose the computed integrals (in hartree, illustrative but chemically reasonable magnitudes for a compact atom) are:

The restricted closed-shell energy formula with $n=2$ orbitals is

Since $K_{11}=J_{11}$ and $K_{22}=J_{22}$ (same-orbital self-exchange cancellation), the diagonal terms reduce to $J_{11}+J_{22}$:

Evaluate term by term.

Summing:

Why the numbers must combine this way: the cross-orbital Coulomb term $J_{12}$ appears with coefficient $2\times 2=4$ (a factor 2 for the off-diagonal double-counting in the sum over $a\neq b$, times the closed-shell factor 2), while the cross-orbital exchange $K_{12}$ appears with coefficient $2\times 1 = 2$ — because only same-spin cross terms exchange, exactly half as many combinatorial pairs as Coulomb. This asymmetry between the coefficients of $J_{12}$ (weight 4) and $K_{12}$ (weight 2) is precisely the "2J−K" structure and is the most common point of formula error in hand calculations — always confirm exchange enters with half the weight of Coulomb for off-diagonal pairs.

Limiting case check: if orbitals 1 and 2 became degenerate and identical ($\psi_2\to\psi_1$), then $J_{12}\to J_{11}$ and $K_{12}\to J_{11}$ (since exchange between an orbital and itself equals its self-Coulomb), and the formula must reduce to four electrons in one orbital: $E \to 4h_1 + 3J_{11}$ (from $2\times2J_{11}-K_{11}\to$ combinatorics of $\binom{4}{2}=6$ pairs total scaled appropriately) — a check readily verified by re-deriving the general $n$-orbital, all-orbitals-identical limit directly from the double sum, which collapses to the single-orbital multi-electron formula used as the seed of Example 1.

Example 3 — Exam/competition level: proving $K_{ii}=J_{ii}$ and locating the self-interaction cancellation explicitly. A common qualifying-exam question: show algebraically, from the definitions of $J_{ij}$ and $K_{ij}$, that setting $i=j$ forces $K_{ii}=J_{ii}$, and explain why this guarantees no unphysical electron self-repulsion survives in $E_{\rm HF}$, even though $\hat J_i$ and $\hat K_i$ both appear explicitly as separate terms inside the Fock operator sum. Start from the general definitions with $i=j$:

Since $\chi_i^*(1)\chi_i(1) = |\chi_i(1)|^2$ and likewise for coordinate 2, the two integrands are term-by-term identical functions of $(1,2)$:

so $K_{ii}=J_{ii}$ is not a numerical coincidence but an algebraic identity — it holds for any normalized orbital $\chi_i$, any nuclear geometry, any basis, because it follows purely from the substitution $j\to i$ in definitions that are structurally identical when both indices coincide.

Consequence for the energy. In the double sum $\tfrac12\sum_{i,j}(J_{ij}-K_{ij})$, the $i=j$ diagonal contributes $\tfrac12\sum_i (J_{ii}-K_{ii}) = \tfrac12\sum_i 0 = 0$. So despite each electron formally "seeing itself" through $\hat J_i$ in the Fock operator sum $\sum_j(\hat J_j - \hat K_j)$, the corresponding self-exchange term $\hat K_i$ exactly and identically removes it when the expectation value is taken — Hartree-Fock theory is therefore rigorously free of one-electron self-interaction error in its energy expression, even though this cancellation is famously not preserved once approximations such as many density-functional exchange-correlation functionals are substituted for exact exchange (a fact often cited as the origin of self-interaction error in DFT, by contrast with exact Hartree-Fock exchange).

Sanity check with the diagonal terms of Example 2: indeed $K_{11}=J_{11}=1.060$ and $K_{22}=J_{22}=0.420$ were assumed exactly in constructing the numbers used — consistent with the identity just proven, and not an independent assumption smuggled in.

Douglas Rayner Hartree was a British mathematical physicist and one of the pioneers of numerical scientific computing, decades before electronic computers existed. Working in the late 1920s, Hartree devised the "self-consistent field" method for atoms — iteratively solving a set of coupled one-electron equations by hand-cranked numerical integration on mechanical calculators (and later, famously, on a differential analyzer he helped build at Manchester). His original scheme used simple products of orbitals, without antisymmetrization, and so lacked exchange terms entirely — an approximation now called the Hartree method, distinct from Hartree-Fock. The crucial correction came from Vladimir Fock (and independently John Slater), who in 1930 showed that Hartree's product wavefunction violated the Pauli exclusion principle, and that replacing it with a properly antisymmetrized determinant introduces the exchange terms derived above. Fock's 1930 paper, written in the language of second quantization and group theory rather than the more accessible determinant notation, established the rigorous theoretical foundation; Slater's nearly simultaneous, more physically transparent treatment using determinants gave the theory its practical form. The combined method — self-consistent orbitals from an antisymmetrized wavefunction — became known as the Hartree-Fock method, arguably the single most consequential approximation scheme in the history of quantum chemistry, underlying essentially every post-HF method (configuration interaction, coupled cluster, Møller-Plesset perturbation theory) as the reference wavefunction from which electron correlation is subsequently added back.