12.09.03 · quantum / identical-particles

Hartree-Fock self-consistent field method

shipped3 tiersLean: none

Anchor (Master): Landau, L. D. & Lifshitz, E. M., *Quantum Mechanics: Non-Relativistic Theory*, 3rd ed. (Pergamon, 1977), Ch. X §§69–70; Szabo & Ostlund, *Modern Quantum Chemistry* (Dover, 1996), Chs. 2–3; Helgaker, T., Jørgensen, P. & Olsen, J., *Molecular Electronic-Structure Theory* (Wiley, 2000), Chs. 10–11

Intuition Beginner

An atom with many electrons is a hard problem because every electron pushes on every other electron at once. You cannot solve for one electron without knowing where all the others are, and you cannot know where the others are without first solving for them. The Hartree-Fock method cuts this knot with a bold simplification: pretend each electron moves alone, not in the field of the other electrons one by one, but in the smooth average field they produce together. One genuine many-body tangle becomes many single-electron problems that talk to each other only through an averaged background.

The catch is that this averaged background depends on the very orbitals you are trying to find. So you guess a set of orbitals, build the average field they create, solve the single-electron problems in that field, and get a new set of orbitals. You repeat until the orbitals you put in match the orbitals you get out. That fixed point is the "self-consistent field." The name says exactly what it does: the field is consistent with the orbitals that generate it.

Why bother? Because this gives a recipe that actually computes. It turns an unsolvable wall of coupled equations into a loop you can run on paper for helium and on a computer for a protein. It is the foundation almost every practical method in quantum chemistry refines or corrects.

Visual Beginner

A flow diagram of the self-consistent loop. Start with a guess for the electron orbitals. From those orbitals, build the average electric field that all the electrons together produce. Put each electron into that field one at a time and solve its single-particle equation, which yields a fresh set of orbitals. Compare the fresh orbitals to the ones you started the loop with. If they disagree, feed the fresh ones back to the top and go around again. If they agree, you have found the self-consistent field and you stop.

The picture captures the essential idea: the field and the orbitals chase each other around a loop until they settle on a pair that match. The arrow that closes the circle is the heart of the method, because it is what makes the field consistent with the particles that create it.

Worked example Beginner

Take helium: two electrons around a nucleus of charge $+ 2$ . We want an estimate of the ground-state energy using the simplest version of the averaging idea.

Step 1. Ignore the repulsion between the two electrons for a moment. Each electron then sees only the nucleus, and the lowest orbital is a hydrogen-like $1 s$ state for nuclear charge $2$ . The energy of one such electron is $- 2^{2} \times 13.6 = - 54.4$ electron-volts, so two electrons give $- 108.8$ electron-volts. That is far too low, because we threw away the repulsion that pushes the energy up.

Step 2. Put the repulsion back in as an average. The two electrons repel, and the average repulsion energy for two $1 s$ electrons of nuclear charge $2$ works out to $+ 34$ electron-volts. Adding this in gives $- 108.8 + 34 = - 74.8$ electron-volts.

Step 3. Compare with experiment. The measured ground-state energy of helium is $- 79.0$ electron-volts. Our averaged estimate of $- 74.8$ is within about five percent.

Step 4. See where the leftover error comes from. We held each electron in a fixed $1 s$ orbital. A better method lets each orbital relax in response to the other electron, which is exactly what the self-consistent loop does, and it recovers most of the remaining gap.

What this tells us: averaging the repulsion turns an unsolvable two-electron problem into an arithmetic one, and it already lands close to the right answer. Letting the orbitals adjust to the averaged field, rather than freezing them, is the improvement that the full method delivers.

Check your understanding Beginner

Exercise (easy, multiple choice).

What makes the Hartree-Fock field "self-consistent"?

A. The energy is exactly equal to the experimental value.
B. The field is computed from the same orbitals it is used to find.
C. The electrons are treated as classical point charges.
D. The nucleus is allowed to move with the electrons.

Hint

Think about the loop: what goes in at the top, and what comes out at the bottom, and what has to match.

Answer

B. The field is computed from the same orbitals it is used to find.

Self-consistency means the orbitals that build the average field are the same orbitals that solving in that field produces. The loop repeats until the input orbitals and the output orbitals agree. It does not require matching experiment, and the electrons remain quantum, not classical.

Formal definition Intermediate+

Consider $N$ electrons with the non-relativistic electronic Hamiltonian, in atomic units, $$ \hat H = \sum_{i=1}^{N} \hat h(i) + \sum_{i<j} \frac{1}{|\mathbf r_i - \mathbf r_j|}, \qquad \hat h = -\tfrac{1}{2}\nabla^2 - \sum_{A} \frac{Z_A}{|\mathbf r - \mathbf R_A|}, $$ where $\hat{h}$ collects the kinetic energy and the electron-nucleus attraction (the one-body part), and the pair sum is the electron-electron repulsion (the two-body part). The state space is the antisymmetric subspace $⋀^{N} H_{1}$ of the $N$ -fold tensor product of the one-electron space $H_{1} = L^{2} (R^{3}) \otimes C^{2}$ , the factor $C^{2}$ carrying spin.

A spin-orbital is a one-particle state $χ_{i} \in H_{1}$ . Given $N$ orthonormal spin-orbitals ${χ_{1}, \dots, χ_{N}}$ , the Slater determinant is the normalised antisymmetrised product $$ \Psi(\mathbf x_1,\dots,\mathbf x_N) = \frac{1}{\sqrt{N!}} \det \big[ \chi_i(\mathbf x_j) \big]_{i,j=1}^{N}, $$ where $x = (r, s)$ bundles space and spin. The determinant changes sign under exchange of any two coordinates and vanishes whenever two spin-orbitals coincide, so it enforces both antisymmetry and the Pauli exclusion principle by construction.

The Hartree-Fock approximation is the restriction of the variational problem to single Slater determinants: minimise the energy expectation value $$ E[\Psi] = \langle \Psi | \hat H | \Psi \rangle $$ over all determinants built from orthonormal spin-orbitals. Using the Slater-Condon rules, the energy of a single determinant is $$ E_{\mathrm{HF}} = \sum_{i=1}^{N} \langle i | \hat h | i \rangle

\frac{1}{2} \sum_{i,j=1}^{N} \big( [ii|jj] - [ij|ji] \big), $w h er e t h e tw o - e l ec t r o nin t e g r a l in c h e mi s t s^{'} n o t a t i o ni s$ [ij|kl] = \iint \chi_i^(\mathbf x_1)\chi_j(\mathbf x_1) \frac{1}{r_{12}} \chi_k^(\mathbf x_2)\chi_l(\mathbf x_2), d\mathbf x_1, d\mathbf x_2. $$ The term $[ii ∣ j j]$ is the Coulomb integral $J_{ij}$ (the classical repulsion of two charge clouds) and $[ij ∣ j i]$ is the exchange integral $K_{ij}$ (a purely quantum term with no classical analogue, surviving only for spin-orbitals of the same spin).

Counterexamples to common slips

The exchange term is not a small correction added by hand. It arises automatically from antisymmetry; dropping it gives the older Hartree method, whose energy $\sum_{i} ⟨ i ∣ \hat{h} ∣ i ⟩ + \frac{1}{2} \sum_{i \neq = j} J_{ij}$ omits $K_{ij}$ and uses orbitals that need not be orthogonal.
The diagonal "self-interaction" terms cancel exactly: at $i = j$ one has $J_{ii} - K_{ii} = 0$ , so an electron does not repel itself. This cancellation is a genuine virtue of Hartree-Fock and fails in approximate density-functional treatments.
Orbital energies $ε_{i}$ are Lagrange multipliers, not the pieces of the total energy. Summing them double-counts the electron-electron interaction: $E_{HF} = \sum_{i} ε_{i} - \frac{1}{2} \sum_{ij} (J_{ij} - K_{ij})$ , not $\sum_{i} ε_{i}$ .

Key theorem with proof Intermediate+

Theorem (Hartree-Fock equations). Let $E_{HF}$ be the energy of a single Slater determinant built from orthonormal spin-orbitals ${χ_{i}}$ . A stationary point of $E_{HF}$ subject to the orthonormality constraints $⟨ χ_{i} ∣ χ_{j} ⟩ = δ_{ij}$ satisfies, after a unitary rotation diagonalising the multiplier matrix, the canonical Hartree-Fock equations $$ \hat f, \chi_i = \big[ \hat h + \hat V_H - \hat K \big], \chi_i = \varepsilon_i, \chi_i, \qquad i = 1,\dots,N, $$ where the local Hartree (direct) potential acts by multiplication, $$ (\hat V_H \chi_i)(\mathbf x_1) = \left[ \sum_{j=1}^{N} \int \frac{|\chi_j(\mathbf x_2)|^2}{r_{12}}, d\mathbf x_2 \right] \chi_i(\mathbf x_1), $$ and the non-local exchange operator acts by an integral kernel, $$ (\hat K \chi_i)(\mathbf x_1) = \sum_{j=1}^{N} \left[ \int \frac{\chi_j^*(\mathbf x_2)\chi_i(\mathbf x_2)}{r_{12}}, d\mathbf x_2 \right] \chi_j(\mathbf x_1). $$

Proof. Form the constrained functional with a Hermitian matrix of Lagrange multipliers $λ_{ij}$ , $$ \mathcal L = E_{\mathrm{HF}}[{\chi_i}] - \sum_{i,j=1}^{N} \lambda_{ij}\big( \langle \chi_i | \chi_j \rangle - \delta_{ij} \big). $$ Vary with respect to $χ_{i}^{*}$ , treating $χ_{i}$ and $χ_{i}^{*}$ as independent (the standard device for complex variations). From the one-body part, $$ \frac{\delta}{\delta \chi_i^} \sum_k \langle k | \hat h | k \rangle = \hat h, \chi_i. $$ For the two-body part, write the interaction energy as $\frac{1}{2} \sum_{k, l} (J_{k l} - K_{k l})$ and differentiate. The variation of $J_{k l} = [k k ∣ l l]$ in $\chi_i^ $co n t r ib u t es a t er m f or e a c ha pp e a r an ceo f in d e x$ i $am o n g$ k,l $; co l l ec t in g b o t han d u s in g t h esy mm e t r y$ J_{kl}=J_{lk}$ gives $$ \frac{\delta}{\delta \chi_i^}, \tfrac12 \sum_{k,l} J_{kl} = \left[ \sum_{j} \int \frac{|\chi_j(\mathbf x_2)|^2}{r_{12}}, d\mathbf x_2 \right] \chi_i(\mathbf x_1) = (\hat V_H \chi_i)(\mathbf x_1). $$ The same bookkeeping for the exchange energy $K_{k l} = [k l ∣ l k]$ yields $$ \frac{\delta}{\delta \chi_i^}, \tfrac12 \sum_{k,l} K_{kl} = \sum_{j} \left[ \int \frac{\chi_j^*(\mathbf x_2)\chi_i(\mathbf x_2)}{r_{12}}, d\mathbf x_2 \right] \chi_j(\mathbf x_1) = (\hat K \chi_i)(\mathbf x_1). $$

The variation of the constraint term is $\sum_{j} λ_{ij} χ_{j}$ . Setting $δ L / δ χ_{i}^{*} = 0$ gives the coupled system $$ \hat f, \chi_i := \big(\hat h + \hat V_H - \hat K\big)\chi_i = \sum_{j} \lambda_{ij}, \chi_j. $$ The operator $\hat{f}$ , the Fock operator, is Hermitian, and it is invariant under any unitary mixing of the occupied spin-orbitals because $\hat{V}_{H}$ and $\hat{K}$ depend only on the projector $\hat{P} = \sum_{j} ∣ χ_{j} ⟩ ⟨ χ_{j} ∣$ onto the occupied space, which a unitary mixing leaves fixed. Choose the mixing that diagonalises the Hermitian multiplier matrix, $λ_{ij} = ε_{i} δ_{ij}$ . In this canonical basis the system decouples into the stated eigenvalue equations $\hat{f} χ_{i} = ε_{i} χ_{i}$ . $□$

Bridge. The Hartree-Fock equations build toward every orbital-based model of matter, and the structural reason is the invariance just used: the Fock operator depends on the occupied orbitals only through the density projector $\hat{P}$ , which is exactly why a stationary point can be presented as an eigenvalue problem rather than a generic coupled system. This is exactly the same density-dependence that makes the equations nonlinear, since $\hat{f}$ is built from the orbitals it acts on, and that nonlinearity is the foundational reason an iterative solution is unavoidable — the central insight of the self-consistent field. The construction generalises in two directions that appear again in later units: replacing the exchange operator by a density functional gives Kohn-Sham theory, and projecting onto a finite basis gives the Roothaan-Hall matrix equations. Putting these together, the single-determinant variational principle identifies the best mean field with the eigenstates of an operator that the mean field itself defines, and that circular definition is the bridge between the abstract minimisation and the concrete computational loop.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Express the total Hartree-Fock energy $E_{HF}$ in terms of the sum of orbital energies $\sum_{i} ε_{i}$ and the two-electron integrals, and explain why $E_{HF} \neq = \sum_{i} ε_{i}$ .

Hint

Take the expectation of $\hat{f} χ_{i} = ε_{i} χ_{i}$ in $χ_{i}$ and sum over $i$ , then compare with the energy formula.

Answer

From $ε_{i} = ⟨ i ∣ \hat{f} ∣ i ⟩ = ⟨ i ∣ \hat{h} ∣ i ⟩ + \sum_{j} (J_{ij} - K_{ij})$ , summing gives $\sum_{i} ε_{i} = \sum_{i} ⟨ i ∣ \hat{h} ∣ i ⟩ + \sum_{ij} (J_{ij} - K_{ij})$ . The energy is $E_{HF} = \sum_{i} ⟨ i ∣ \hat{h} ∣ i ⟩ + \frac{1}{2} \sum_{ij} (J_{ij} - K_{ij})$ . Therefore $E_{HF} = \sum_{i} ε_{i} - \frac{1}{2} \sum_{ij} (J_{ij} - K_{ij})$ . The sum of orbital energies counts each electron-electron interaction twice, so subtracting one factor of the interaction corrects the double counting. Rubric: full credit for the relation and the double-counting explanation.

Exercise 4 (medium, short-answer).

Show that the Fock operator $\hat{f} = \hat{h} + \hat{V}_{H} - \hat{K}$ is invariant under a unitary mixing $χ_{i} \mapsto \sum_{j} U_{j i} χ_{j}$ of the occupied spin-orbitals, with $U$ unitary.

Hint

Show that $\hat{V}_{H}$ and $\hat{K}$ depend on the orbitals only through the projector $\hat{P} = \sum_{i} ∣ χ_{i} ⟩ ⟨ χ_{i} ∣$ .

Answer

The Hartree potential has kernel $\sum_{j} χ_{j} (x_{2}) χ_{j}^{*} (x_{2}) = ρ (x_{2})$ , the diagonal of $\hat{P}$ , and the exchange operator has kernel $\sum_{j} χ_{j} (x_{1}) χ_{j}^{*} (x_{2}) / r_{12} = P (x_{1}, x_{2}) / r_{12}$ , where $P (x_{1}, x_{2})$ is the kernel of $\hat{P}$ . Under $χ_{i} \mapsto \sum_{j} U_{j i} χ_{j}$ with $U$ unitary, $\hat{P} = \sum_{i} ∣ χ_{i} ⟩ ⟨ χ_{i} ∣$ is unchanged because $\sum_{i} U_{j i} U_{k i}^{*} = δ_{j k}$ . Since both $\hat{V}_{H}$ and $\hat{K}$ are functions of $\hat{P}$ alone, and $\hat{h}$ does not involve the orbitals, $\hat{f}$ is invariant. Rubric: full credit for reducing both terms to $\hat{P}$ and using unitarity.

Exercise 5 (medium, short-answer).

State Koopmans' theorem and explain in one sentence why it provides an estimate of the first ionisation potential.

Hint

Remove an electron from orbital $a$ while holding the remaining orbitals fixed (frozen-orbital approximation) and compute the energy change.

Answer

Koopmans' theorem states that, within the frozen-orbital approximation, the ionisation potential for removing an electron from occupied spin-orbital $a$ is $- ε_{a}$ , and the electron affinity for adding an electron to virtual spin-orbital $r$ is $- ε_{r}$ . It estimates the first ionisation potential because the energy difference between the $N$ -electron determinant and the $(N - 1)$ -electron determinant with orbital $a$ vacated, with all other orbitals unrelaxed, equals exactly $- ε_{a}$ . Rubric: full credit for the statement plus the frozen-orbital justification.

Exercise 7 (hard, short-answer).

Derive the Roothaan-Hall matrix equations $F C = S C ε$ by expanding each spatial orbital in a finite basis $ψ_{i} = \sum_{μ} C_{μ i} ϕ_{μ}$ , and identify why the overlap matrix $S$ appears.

Hint

Substitute the expansion into $\hat{f} ψ_{i} = ε_{i} ψ_{i}$ , left-multiply by $ϕ_{ν}^{*}$ , and integrate. The basis ${ϕ_{μ}}$ need not be orthonormal.

Answer

Substituting $ψ_{i} = \sum_{μ} C_{μ i} ϕ_{μ}$ into $\hat{f} ψ_{i} = ε_{i} ψ_{i}$ , projecting onto $ϕ_{ν}$ , and integrating gives $\sum_{μ} ⟨ ϕ_{ν} ∣ \hat{f} ∣ ϕ_{μ} ⟩ C_{μ i} = ε_{i} \sum_{μ} ⟨ ϕ_{ν} ∣ ϕ_{μ} ⟩ C_{μ i}$ . Defining the Fock matrix $F_{ν μ} = ⟨ ϕ_{ν} ∣ \hat{f} ∣ ϕ_{μ} ⟩$ and the overlap matrix $S_{ν μ} = ⟨ ϕ_{ν} ∣ ϕ_{μ} ⟩$ , this reads $\sum_{μ} F_{ν μ} C_{μ i} = ε_{i} \sum_{μ} S_{ν μ} C_{μ i}$ , that is, $FC = SC ε$ with $ε$ diagonal. The overlap matrix $S$ appears precisely because a general basis is non-orthonormal; it reduces to the identity for an orthonormal basis, recovering an ordinary eigenvalue problem. Because $F$ depends on $C$ through the density matrix $P_{μν} = 2 \sum_{i} C_{μ i} C_{ν i}^{*}$ , the equation is nonlinear and must be solved self-consistently. Rubric: full credit for the projection derivation, the role of $S$ , and noting the nonlinearity.

Exercise 8 (hard, short-answer).

Explain why the Hartree-Fock energy, despite being variational, leaves a residual error called the correlation energy, and give the standard definition of that quantity.

Hint

The trial space is restricted to single determinants. Ask what the exact wavefunction can do that a single determinant cannot.

Answer

A single Slater determinant builds in exchange correlation (same-spin electrons avoid one another through antisymmetry) but treats the remaining motion as that of independent particles in a mean field. The exact wavefunction has instantaneous correlation between opposite-spin electrons, captured only by mixing in additional determinants. The variational principle guarantees $E_{HF} \geq E_{exact}$ , and the gap is the correlation energy $E_{corr} = E_{exact} - E_{HF}$ (a negative number), defined in the limit of a complete basis to separate it from basis-set error. Post-Hartree-Fock methods such as configuration interaction, coupled cluster, and many-body perturbation theory recover $E_{corr}$ by expanding the trial space beyond a single determinant. Rubric: full credit for the independent-particle limitation plus the definition $E_{corr} = E_{exact} - E_{HF}$ .

Advanced results Master

Theorem (Koopmans, 1934). Within the frozen-orbital approximation — the occupied spin-orbitals of the $N$ - and $(N - 1)$ -electron determinants are held identical — the ionisation potential for removing an electron from canonical occupied spin-orbital $a$ equals $- ε_{a}$ , and the electron affinity for filling canonical virtual spin-orbital $r$ equals $- ε_{r}$ . The proof computes $E_{N} - E_{N - 1}^{(a)}$ directly from the Slater-Condon energy expression; all one- and two-electron terms not involving $a$ cancel between the two determinants, and the surviving terms assemble into $⟨ a ∣ \hat{h} ∣ a ⟩ + \sum_{j \neq = a} (J_{aj} - K_{aj}) = ε_{a}$ . The result is an approximation by two compensating errors: it neglects orbital relaxation of the cation (which lowers $E_{N - 1}$ , overestimating the ionisation potential) and it neglects the differing correlation energies of the two systems (which acts in the opposite direction). For valence ionisations of closed-shell molecules the two errors partially cancel, and Koopmans estimates are useful to within an electron-volt.

Theorem (Brillouin, 1934). The Hartree-Fock determinant $Ψ_{0}$ does not couple to any singly excited determinant $Ψ_{i}^{a}$ through the Hamiltonian: $⟨ Ψ_{0} ∣ \hat{H} ∣ Ψ_{i}^{a} ⟩ = 0$ for every occupied $i$ and virtual $a$ . This is the matrix element $⟨ a ∣ \hat{f} ∣ i ⟩$ , which vanishes because the canonical orbitals diagonalise $\hat{f}$ and occupied and virtual labels are distinct. Brillouin's theorem is the statement that the Hartree-Fock determinant is stationary against single excitations, which is the variational condition itself, and it explains why the leading correlation correction in configuration interaction comes from double, not single, excitations.

Self-consistent-field iteration. The nonlinear eigenproblem $\hat{f} [{χ}] χ_{i} = ε_{i} χ_{i}$ is solved as a fixed point. Given a density projector $\hat{P}^{(k)}$ , build $\hat{f}^{(k)} = \hat{h} + \hat{V}_{H} [\hat{P}^{(k)}] - \hat{K} [\hat{P}^{(k)}]$ , diagonalise it, occupy the $N$ lowest spin-orbitals (the aufbau prescription), and form $\hat{P}^{(k + 1)}$ . Iterate until $∥ \hat{P}^{(k + 1)} - \hat{P}^{(k)} ∥$ falls below tolerance. Plain iteration can oscillate or diverge; in practice it is stabilised by density mixing or, most commonly, by Pulay's direct inversion in the iterative subspace (DIIS), which extrapolates a new density from a least-squares combination of previous error vectors built from the commutator $[\hat{f}, \hat{P}]$ , whose vanishing is equivalent to self-consistency.

Restricted and unrestricted formulations. For a closed-shell system, restricted Hartree-Fock (RHF) forces each spatial orbital to hold one spin-up and one spin-down electron, halving the number of independent orbitals and giving a spin-pure singlet. For open-shell systems, unrestricted Hartree-Fock (UHF) lets spin-up and spin-down spatial orbitals differ, gaining variational freedom at the cost of spin contamination, since the UHF determinant is not generally an eigenstate of $\hat{S}^{2}$ . The two coincide for closed shells away from the regime where RHF becomes unstable.

The Hartree limit and its defect. Dropping $\hat{K}$ returns the Hartree equations, in which each electron moves in the classical averaged charge density of the others. The Hartree wavefunction is a plain product, not antisymmetrised, so it violates the Pauli principle and suffers an uncancelled self-interaction $J_{ii} \neq = 0$ . Fock's 1930 antisymmetrisation supplies exactly the exchange term that cancels the self-interaction and enforces the exclusion principle, which is why the Hartree-Fock energy of a closed shell is strictly below the Hartree energy.

Stability and symmetry breaking. A Hartree-Fock stationary point is a minimum only if the orbital Hessian (the electronic Hessian, whose blocks are the response of $\hat{f}$ to orbital rotations) is positive definite. When it develops a negative eigenvalue, the symmetric solution is a saddle and a lower-energy symmetry-broken solution exists, the molecular analogue of the Stoner instability. This is the mean-field shadow of strong correlation: where Hartree-Fock breaks a symmetry of the exact Hamiltonian, the single-determinant ansatz is signalling its own inadequacy.

Synthesis. The Hartree-Fock construction is the foundational reason the independent-particle picture survives into a genuinely interacting many-electron world: the central insight is that capping the variational search at single determinants converts the two-body Hamiltonian into a one-body Fock operator whose own definition depends on its eigenvectors, and this is exactly the self-consistency that the Hartree limit, the exchange correction, and the Roothaan matrix form all inherit. The same density-projector structure that makes the Fock operator orbital-rotation invariant generalises directly: replacing $- \hat{K}$ by a local exchange-correlation potential identifies the Hartree-Fock equations with the Kohn-Sham equations of density-functional theory, and projecting onto a finite basis identifies them with the Roothaan-Hall generalised eigenproblem. Putting these together, Koopmans' theorem reads the orbital energies as ionisation data, Brillouin's theorem certifies stationarity against single excitations, and the correlation energy measures precisely what the single determinant cannot reach — the three of them triangulate the exact boundary of the mean-field approximation. The bridge from this unit to the rest of electronic-structure theory is that every post-Hartree-Fock method is defined as a correction to this determinant, so the self-consistent field is not merely one method among many but the reference state against which the others are calibrated.

Full proof set Master

Proposition (variational lower bound and the Hartree-Fock inequality). Let $\hat{H}$ be the electronic Hamiltonian with ground-state energy $E_{0}$ , and let $S$ be the set of normalised single Slater determinants. Then $$ E_0 ;\le; E_{\mathrm{HF}} := \min_{\Psi\in\mathcal S} \langle\Psi|\hat H|\Psi\rangle, $$ and the minimising determinant exists when $\hat{H}$ is bounded below with discrete low-lying spectrum.

Proof. Every normalised state, determinant or not, satisfies the Rayleigh-Ritz inequality $⟨ Ψ∣ \hat{H} ∣Ψ ⟩ \geq E_{0}$ , since expanding $Ψ$ in exact eigenstates $Φ_{n}$ with energies $E_{n} \geq E_{0}$ gives $⟨ Ψ∣ \hat{H} ∣Ψ ⟩ = \sum_{n} ∣ c_{n} ∣^{2} E_{n} \geq E_{0} \sum_{n} ∣ c_{n} ∣^{2} = E_{0}$ . The Slater determinants form a subset of the normalised antisymmetric states, so the infimum of the energy over $S$ is bounded below by $E_{0}$ , giving $E_{HF} \geq E_{0}$ . Existence of the minimiser is the content of the Lieb-Simon theorem: for the molecular Hamiltonian with total nuclear charge exceeding the electron number, the Hartree-Fock functional attains its minimum on the manifold of orthonormal orbital frames, because the functional is weakly lower semicontinuous and coercive on that manifold once the negative ions are excluded. The minimiser is a stationary point and therefore satisfies the canonical Hartree-Fock equations of the Key-theorem section. $□$

Proposition (the canonical condition $[\hat{f}, \hat{P}] = 0$ ). At a Hartree-Fock stationary point the Fock operator commutes with the density projector $\hat{P} = \sum_{i = 1}^{N} ∣ χ_{i} ⟩ ⟨ χ_{i} ∣$ .

Proof. The off-diagonal multiplier equation $\hat{f} χ_{i} = \sum_{j} λ_{ij} χ_{j}$ says that $\hat{f}$ maps the occupied space into itself, so $\hat{P}^{⊥} \hat{f} \hat{P} = 0$ , where $\hat{P}^{⊥} = 1 - \hat{P}$ . Taking the adjoint, and using that $\hat{f}$ and the projectors are Hermitian, gives $\hat{P} \hat{f} \hat{P}^{⊥} = 0$ . Then $$ [\hat f, \hat P] = \hat f\hat P - \hat P\hat f = (\hat P + \hat P^\perp)\hat f\hat P - \hat P\hat f(\hat P + \hat P^\perp) = \hat P^\perp \hat f\hat P - \hat P\hat f\hat P^\perp = 0. $$ Conversely, if $[\hat{f}, \hat{P}] = 0$ then $\hat{f}$ preserves the occupied space, which is the stationarity condition. The commutator $[\hat{f}, \hat{P}]$ is therefore the natural residual measuring departure from self-consistency, and driving it to zero is the target of the iteration. $□$

Proposition (Koopmans' identity for the orbital energy). In the canonical basis, $ε_{a} = ⟨ a ∣ \hat{h} ∣ a ⟩ + \sum_{j = 1}^{N} (J_{aj} - K_{aj})$ , and the frozen-orbital energy difference $E_{N} - E_{N - 1}^{(a)}$ equals $ε_{a}$ .

Proof. Diagonal of the Fock eigenvalue equation: $ε_{a} = ⟨ a ∣ \hat{f} ∣ a ⟩ = ⟨ a ∣ \hat{h} ∣ a ⟩ + ⟨ a ∣ \hat{V}_{H} ∣ a ⟩ - ⟨ a ∣ \hat{K} ∣ a ⟩$ , and inserting the kernels gives $⟨ a ∣ \hat{V}_{H} ∣ a ⟩ = \sum_{j} J_{aj}$ and $⟨ a ∣ \hat{K} ∣ a ⟩ = \sum_{j} K_{aj}$ , hence $ε_{a} = ⟨ a ∣ \hat{h} ∣ a ⟩ + \sum_{j} (J_{aj} - K_{aj})$ . Now write the $N$ -electron energy $E_{N} = \sum_{i} ⟨ i ∣ \hat{h} ∣ i ⟩ + \frac{1}{2} \sum_{ij} (J_{ij} - K_{ij})$ and the frozen $(N - 1)$ -electron energy $E_{N - 1}^{(a)}$ with the sum restricted to $i, j \neq = a$ . Subtracting, the terms with neither index equal to $a$ cancel, and the terms with exactly one index equal to $a$ appear twice in $E_{N}$ at half weight, assembling to $⟨ a ∣ \hat{h} ∣ a ⟩ + \sum_{j \neq = a} (J_{aj} - K_{aj})$ . Since $J_{aa} - K_{aa} = 0$ , this equals $ε_{a}$ . $□$

Proposition (Brillouin's theorem). $⟨ Ψ_{0} ∣ \hat{H} ∣ Ψ_{i}^{a} ⟩ = 0$ for every singly excited determinant $Ψ_{i}^{a}$ obtained from the Hartree-Fock determinant $Ψ_{0}$ by promoting occupied $i$ to virtual $a$ .

Proof. By the Slater-Condon rule for determinants differing in one spin-orbital, $⟨ Ψ_{0} ∣ \hat{H} ∣ Ψ_{i}^{a} ⟩ = ⟨ a ∣ \hat{h} ∣ i ⟩ + \sum_{j \in occ} ([ai ∣ j j] - [aj ∣ j i]) = ⟨ a ∣ \hat{f} ∣ i ⟩$ . In the canonical basis $\hat{f}$ is diagonal, so $⟨ a ∣ \hat{f} ∣ i ⟩ = ε_{i} ⟨ a ∣ i ⟩ = 0$ because the occupied $i$ and virtual $a$ are distinct orthonormal eigenvectors. $□$

Connections Master

Fermionic Fock space, Pauli exclusion, and anticommutators 12.13.02. The Slater determinant is the configuration-space face of a single occupied state in fermionic Fock space; antisymmetry under exchange, the vanishing of a determinant with a repeated orbital, and the canonical anticommutation relations are three presentations of the same exclusion principle, and the Hartree-Fock determinant is precisely a product of creation operators acting on the vacuum. The exchange operator $\hat{K}$ is the one-body shadow of the anticommutator structure of the two-body interaction.
Time-independent perturbation theory 12.07.01. Hartree-Fock is the natural zeroth-order reference for many-body perturbation theory: Møller-Plesset perturbation theory partitions $\hat{H} = \hat{f}_{sum} + (\hat{H} - \hat{f}_{sum})$ , with the sum of Fock operators as the unperturbed Hamiltonian and the fluctuation potential as the perturbation, and Brillouin's theorem is exactly what makes the first correction to the energy vanish so that correlation enters at second order.
Operators, observables, and Hermiticity 12.02.02. The Fock operator is a self-adjoint one-body operator whose eigenvalues are the orbital energies and whose eigenvectors are the canonical orbitals; the entire method rests on the spectral theorem for Hermitian operators applied to an operator that is itself a functional of its own occupied eigenspace, the self-referential structure that distinguishes mean-field theory from ordinary linear eigenproblems.
Hilbert-space formalism 12.02.01. The antisymmetric $N$ -electron state space is the $N$ -th exterior power of the one-particle Hilbert space, and the variational principle is the Rayleigh-Ritz minimisation restricted to decomposable wedge tensors; the inner-product geometry that defines orthonormality of spin-orbitals and the overlap matrix in a non-orthonormal basis is the same Hilbert-space structure that underlies every expectation value in the theory.

Historical & philosophical context Master

Douglas Hartree introduced the self-consistent field in two 1928 papers in the Proceedings of the Cambridge Philosophical Society ^{[Hartree 1928]}, building on his father William Hartree's facility with numerical computation. Hartree's electrons each moved in the spherically averaged field of the others, and he iterated the orbitals by hand until the input and output fields matched, coining the phrase "self-consistent field" for the converged result. The wavefunction was a simple product of one-electron orbitals, which respected neither antisymmetry nor the exclusion principle as a property of the trial function — these were imposed afterward by the aufbau filling of states.

Vladimir Fock supplied the missing ingredient in 1930 in Zeitschrift für Physik ^{[Fock 1930]}, replacing Hartree's product by an antisymmetrised wavefunction and deriving the additional exchange term that the antisymmetry forces. Independently, John Slater had introduced the determinantal many-electron wavefunction in his 1929 Physical Review paper on complex spectra ^{[Slater 1929]}, where the determinant form made the antisymmetry manifest and led to the systematic Slater-Condon rules for matrix elements. The synthesis — Slater's determinant as the trial function, Fock's variational treatment producing the exchange operator — is the method now called Hartree-Fock, with the determinant frequently named for Slater.

The theory acquired its computational form through Clemens Roothaan's 1951 Reviews of Modern Physics paper ^{[Roothaan 1951]}, which recast the integro-differential equations as the matrix generalised eigenproblem $FC = SC ε$ in a finite basis, the form implemented in essentially every electronic-structure program since. Tjalling Koopmans' 1934 theorem connecting orbital energies to ionisation potentials, and Léon Brillouin's contemporaneous result on the vanishing of single-excitation matrix elements, fixed the interpretive scaffolding. Landau and Lifshitz present the atomic self-consistent field compactly in §§69–70 of Quantum Mechanics, deriving the Hartree and Hartree-Fock equations from the variational principle and emphasising the exchange term as the non-local correction that distinguishes the antisymmetrised treatment from Hartree's original product ansatz ^{[Landau-Lifshitz 1977]}.

Bibliography Master

@article{Hartree1928,
  author  = {Hartree, Douglas R.},
  title   = {The Wave Mechanics of an Atom with a Non-Coulomb Central Field. Part I. Theory and Methods},
  journal = {Proceedings of the Cambridge Philosophical Society},
  volume  = {24},
  number  = {1},
  year    = {1928},
  pages   = {89--110}
}

@article{Fock1930,
  author  = {Fock, Vladimir},
  title   = {N{\"a}herungsmethode zur L{\"o}sung des quantenmechanischen Mehrk{\"o}rperproblems},
  journal = {Zeitschrift f{\"u}r Physik},
  volume  = {61},
  number  = {1--2},
  year    = {1930},
  pages   = {126--148}
}

@article{Slater1929,
  author  = {Slater, John C.},
  title   = {The Theory of Complex Spectra},
  journal = {Physical Review},
  volume  = {34},
  number  = {10},
  year    = {1929},
  pages   = {1293--1322}
}

@article{Roothaan1951,
  author  = {Roothaan, Clemens C. J.},
  title   = {New Developments in Molecular Orbital Theory},
  journal = {Reviews of Modern Physics},
  volume  = {23},
  number  = {2},
  year    = {1951},
  pages   = {69--89}
}

@article{Koopmans1934,
  author  = {Koopmans, Tjalling},
  title   = {{\"U}ber die Zuordnung von Wellenfunktionen und Eigenwerten zu den einzelnen Elektronen eines Atoms},
  journal = {Physica},
  volume  = {1},
  number  = {1--6},
  year    = {1934},
  pages   = {104--113}
}

@book{LandauLifshitzQM1977,
  author    = {Landau, Lev D. and Lifshitz, Evgeny M.},
  title     = {Quantum Mechanics: Non-Relativistic Theory},
  edition   = {3},
  publisher = {Pergamon Press},
  year      = {1977},
  note      = {Course of Theoretical Physics, Vol. 3, Ch. X}
}

@book{SzaboOstlund1996,
  author    = {Szabo, Attila and Ostlund, Neil S.},
  title     = {Modern Quantum Chemistry: Introduction to Advanced Electronic Structure Theory},
  publisher = {Dover Publications},
  year      = {1996}
}

@book{SlaterAtomicStructure1960,
  author    = {Slater, John C.},
  title     = {Quantum Theory of Atomic Structure, Vol. I},
  publisher = {McGraw-Hill},
  year      = {1960}
}

@book{HelgakerJorgensenOlsen2000,
  author    = {Helgaker, Trygve and J{\o}rgensen, Poul and Olsen, Jeppe},
  title     = {Molecular Electronic-Structure Theory},
  publisher = {Wiley},
  year      = {2000}
}

@article{LiebSimon1977,
  author  = {Lieb, Elliott H. and Simon, Barry},
  title   = {The Hartree-Fock theory for Coulomb systems},
  journal = {Communications in Mathematical Physics},
  volume  = {53},
  number  = {3},
  year    = {1977},
  pages   = {185--194}
}

@article{Pulay1980,
  author  = {Pulay, P{\'e}ter},
  title   = {Convergence acceleration of iterative sequences. The case of {SCF} iteration},
  journal = {Chemical Physics Letters},
  volume  = {73},
  number  = {2},
  year    = {1980},
  pages   = {393--398}
}

Prerequisites

12.02.01
12.02.02
12.07.01
12.13.02

Tier anchors

beginner: Griffiths, *Introduction to Quantum Mechanics*, 2nd ed. (Pearson, 2005), Ch. 7 (variational principle) and Ch. 5 (multi-electron atoms)
intermediate: Szabo, A. & Ostlund, N. S., *Modern Quantum Chemistry* (Dover, 1996), Ch. 3 (the Hartree-Fock approximation); Slater, J. C., *Quantum Theory of Atomic Structure*, Vol. I (McGraw-Hill, 1960)
master: Landau, L. D. & Lifshitz, E. M., *Quantum Mechanics: Non-Relativistic Theory*, 3rd ed. (Pergamon, 1977), Ch. X §§69–70; Szabo & Ostlund, *Modern Quantum Chemistry* (Dover, 1996), Chs. 2–3; Helgaker, T., Jørgensen, P. & Olsen, J., *Molecular Electronic-Structure Theory* (Wiley, 2000), Chs. 10–11

References

Landau, L. D. & Lifshitz, E. M. — Quantum Mechanics: Non-Relativistic Theory · 3rd ed. (Pergamon, 1977), Ch. X §§69–70 (the self-consistent field; Hartree and Hartree-Fock equations for the atom)
Hartree, D. R. — The Wave Mechanics of an Atom with a Non-Coulomb Central Field · Proc. Cambridge Philos. Soc. 24, 89–110 and 111–132 (1928) (Parts I and II — the self-consistent field method)
Fock, V. — Näherungsmethode zur Lösung des quantenmechanischen Mehrkörperproblems · Zeitschrift für Physik 61, 126–148 (1930) (antisymmetrised wavefunction and the exchange term)
Slater, J. C. — The Theory of Complex Spectra · Physical Review 34, 1293–1322 (1929) (the determinantal many-electron wavefunction; Slater-Condon rules)
Roothaan, C. C. J. — New Developments in Molecular Orbital Theory · Reviews of Modern Physics 23, 69–89 (1951) (the Roothaan-Hall matrix form of the Hartree-Fock equations in a finite basis)
Szabo, A. & Ostlund, N. S. — Modern Quantum Chemistry · Dover (1996), Chs. 2–3 (Slater determinants, Slater-Condon rules, Hartree-Fock and Roothaan equations)

Estimated time

beginner: 18m
intermediate: 45m
master: 80m