← quantum · contents

§ Phase ix · Pauli & The Periodic Table · Chapter 02

Hartree-Fock

By 1928 the Schrödinger equation could be written down for any atom in the periodic table, but it had only ever been solved for hydrogen. A Manchester physicist named Douglas Hartree decided to brute-force the rest with paper, pencil, and a mechanical differential analyzer, by pretending every electron felt only the average pull of all the others. Two years later a Leningrad theorist named Vladimir Fock added the one piece Hartree had missed, the antisymmetry that Pauli demanded. The workhorse of modern chemistry was born, and it is still running today inside every pharmaceutical pipeline on Earth.

In the spring of 1926, Erwin Schrödinger published a wave equation that, in principle, contained all of chemistry. Solve it for the electrons in any atom or molecule, and you could read off the energy levels, the colors, the bond lengths, the entire periodic table. There was only one problem. The equation could be solved exactly only for the simplest atom in the universe, hydrogen, which has a single electron and a single proton. For helium with two electrons, for lithium with three, for carbon with six, for uranium with ninety-two, the equation was a tangle of coupled coordinates that no human and no machine in 1928 had any hope of cracking.

The difficulty is easy to see. For a single electron you have three coordinates (x, y, z) and one wavefunction ψ(x, y, z) you must find. For two electrons you have six coordinates, three for each particle, and one wavefunction ψ(x₁, y₁, z₁, x₂, y₂, z₂) defined over that six-dimensional space. For N electrons you have a function in 3N dimensions, and each pair of electrons repels every other pair through a Coulomb term that couples all the coordinates together. The number of grid points you would need to tabulate such a function blows up exponentially. Treating a single iron atom (26 electrons, 78 dimensions) at the laziest possible 10 grid points per axis would demand 10⁷⁸ numbers, more entries than there are atoms in the observable universe.

The man who decided this was no excuse to give up was a quiet, methodical Englishman with a deep affection for machinery. Douglas Hartree had trained as an artillery computer in the First World War, worked out trajectories for naval shells, and developed a lifelong faith that any problem could be reduced to a procedure if you were patient enough about it. In 1927, he proposed a procedure for the many-electron Schrödinger equation. He called it the self-consistent field method. Most of his colleagues thought it was a bit of a swindle, until it began producing answers that matched experiment.

Hartree’s trick was to refuse to look at the full N-electron wavefunction at all. Instead he asked a smaller question. Imagine you are a single electron in a heavy atom. You feel the pull of the nucleus, and you feel the push of every other electron. You cannot track each of those other electrons individually, there are too many. But you do not need to. From your point of view, the other electrons are smeared out into a cloud of negative charge that screens the nucleus. What you really feel is the bare nuclear potential plus the average electrostatic potential of that cloud. If you knew the cloud, you could solve a simple one-electron Schrödinger equation for your own orbital. The trouble is that you, too, are part of the cloud that every other electron sees. So all the orbitals must be solved together, each consistent with the average field produced by all the others.

This is a chicken-and-egg problem, and Hartree’s answer was to break the loop by guessing. Start with a crude trial set of orbitals, perhaps hydrogenic ones with effective charges. Compute the cloud they produce. Use that cloud to write down a one-electron equation for each orbital. Solve those equations. You now have a new set of orbitals. If they look like the old ones, you have converged. If they do not, throw the old ones away, use the new ones to compute a new cloud, and try again. Keep going until the input matches the output. The fixed point of this iteration is the self-consistent field, and the orbitals at that fixed point are Hartree’s approximate answer to the many-body problem.

By 1929 Hartree’s method had been published, debated, and quietly criticized. The criticism was sharp and it came in two flavors. First, mathematicians like John Slater pointed out that Hartree had pulled his one-electron equations out of physical intuition rather than from a real variational principle. Second, and more seriously, Hartree’s product wavefunction was wrong in a way that Wolfgang Pauli had made famous in 1925. Pauli had insisted that any wavefunction for identical fermions must change sign when you swap any two of them. Hartree’s wavefunction (a simple product of one-electron orbitals) does no such thing. Swap two electrons in a Hartree product and you get exactly the same function back. The Pauli principle, which is the law of the periodic table itself, is silently violated at the root of the method.

In 1930, a young Soviet theorist in Leningrad fixed this. His name was Vladimir Fock. He was twenty-two years younger than Hartree, fiercely mathematical, and deeply versed in the new group-theoretical language that Hermann Weyl and Eugene Wigner had brought to quantum mechanics. Fock saw that the right way to enforce antisymmetry was not a simple product but a Slater determinant, a kind of generalized determinant whose rows are labeled by orbitals and whose columns are labeled by electrons. Swap any two columns of a determinant, and you flip its sign. The Pauli principle is built into the algebra.

Once the wavefunction is antisymmetric, the variational principle (minimize the expectation value of the energy) produces a new set of one-electron equations that look like Hartree’s but with one extra term. That extra term is called the exchange interaction. It has no classical analogue. It arises purely from the antisymmetry of the wavefunction, from the way electrons can interfere with their own swapped copies. For electrons of the same spin, the exchange term reduces the effective Coulomb repulsion, because two same-spin electrons are forbidden from being at the same point and so naturally stay further apart than two distinguishable particles would. This is the famous Fermi hole that surrounds every electron in a Hartree-Fock calculation, a small bubble of suppressed same-spin density carved out by Pauli rather than by force.

The arithmetic of one Hartree-Fock cycle is not, in itself, terrible. The Fock operator is a one-electron object: a kinetic energy term, a nuclear attraction, a Coulomb repulsion built from the current density, and the exchange term that Fock added. You solve a one-electron Schrödinger-like equation in this effective potential and read off the new orbitals. The trouble is the iteration. Each cycle takes the new orbitals as input to build a new potential, and the loop must continue until the potential stops moving. For a small atom this might mean a dozen cycles. For a medium-sized molecule with a few hundred basis functions it might mean a hundred, with each cycle scaling as the fourth power of the basis-set size. The reason Hartree-Fock did not take over chemistry until the 1960s is simply that the world did not have the computers to feed it. By the time IBM mainframes were widely available, the algorithm was waiting.

⟨Ψ|Ĥ|Ψ⟩ = Σ ⟨ψᵢ|ĥ|ψᵢ⟩ + (1/2) Σ ( ⟨ψᵢψⱼ‖ψᵢψⱼ⟩ − ⟨ψᵢψⱼ‖ψⱼψᵢ⟩ )

F̂ ψᵢ = εᵢ ψᵢ

By the 1960s, Hartree-Fock had become the default starting point for any serious electronic-structure calculation, and the post-Hartree-Fock industry had begun in earnest. Configuration interaction (CI) takes the HF Slater determinant and adds excited determinants to the mix, letting the wavefunction relax beyond a single configuration. Coupled cluster theory, invented by Fritz Coester and Hermann Kümmel in nuclear physics and imported to chemistry by Jiří Čížek in 1966, organizes those excitations in an exponential ansatz that scales much more gently with system size. Density functional theory, which Walter Kohn and Lu Jeu Sham introduced in 1965, sidesteps the wavefunction altogether and works directly with the electron density, using a Kohn-Sham orbital picture that inherits the self-consistent iteration straight from Hartree-Fock. Every one of these methods owes its computational shape to the Manchester physicist who decided in 1927 that brute-forcing the periodic table was a reasonable thing to do.

The cultural footprint is hard to overstate. When a pharmaceutical company designs a new kinase inhibitor in 2026, the chemists in front of the screen are running a coupled-cluster calculation on a candidate molecule with a Hartree-Fock reference, in software whose data flow is recognizably the Hartree-Fock loop. When a battery company simulates lithium intercalation in a new cathode material, they are solving Kohn-Sham equations whose orbital eigenproblem inherits its structure from Fock. When the JWST team models the spectrum of a methane band in an exoplanet’s atmosphere, they use line lists computed against HF-based potential energy surfaces. The 99% of the energy that Hartree and Fock could reach with paper and brass turns out to be the foundation that every more accurate method builds on.

It is worth remembering how the story began, because the lesson is not really about chemistry. It is about how to handle a problem that has no analytic solution and no hope of one. The right move, when N-body coupling is hopeless, is to factor the problem: replace it with N copies of a one-body problem in a self-consistent average field, then iterate until the average is consistent with the one-body answers. Hartree’s procedure was the first appearance of this idea in modern physics, but it would reappear in dozens of disguises over the next ninety years: in the BCS theory of superconductivity, in nuclear shell-model calculations, in plasma physics, in the cosmological density-field equations that simulate the universe at large scales. Anywhere you cannot afford to track every particle, you replace each particle with a one-body problem in the average field of all the others and iterate. The pattern is the same. Only the field changes.

It is also worth saying a word about the temperaments behind the method, because they shape the story too. Hartree was a maker. He built his own mechanical computers, soldered his own circuits, and taught a generation of British students that numerical analysis was a respectable physics. He spent the Second World War running the differential analyzer on ballistic problems for the Ministry of Supply and emerged in 1946 to find that the new electronic computers in the United States, the ENIAC and its descendants, were doing in seconds what his brass disks took weeks to do. Rather than mourn his obsolete machine he flew to Philadelphia, learned to program ENIAC, came home, and helped bring digital computing to British universities. The first electronic Hartree-Fock calculation in Britain was run on the Manchester Mark 1 in 1950, on a programme partly drafted by Hartree’s own hand. Fock, by contrast, was a deeply theoretical creature who spent most of his career inside the rigorous formalism of Hilbert spaces and group representations, surviving the political turbulence of Stalin-era Leningrad with a quiet stubbornness and a refusal to do applied work that he considered beneath the dignity of the mathematics. He never wrote a line of code in his life. Yet between the two of them, the chemist’s bench and the mathematician’s office, they handed the next two generations a method that would consume more CPU time than any other algorithm in chemistry.