Lecture 5 · Part I: Quantum mechanics you need
Pauli exclusion and Slater determinants
What makes many electrons harder than one, and what does the Pauli principle demand?
You will be able to explain
- why electron–electron repulsion makes the many-electron problem hard
- what spin is, and what a spin-orbital is
- why electron wavefunctions are antisymmetric, and how a Slater determinant builds that in
- how antisymmetry gives the Pauli exclusion principle and the exchange hole
- why codes store one-electron orbitals and the density instead of the many-electron wavefunction
By Osman Goni Ridwan · Last reviewed
DFT & VASP Course · Lecture 51 / 16
Lectures 2 to 4 were mostly about one electron at a time. Real atoms, molecules and solids have many, and they repel each other and obey a rule that has no classical counterpart: two electrons can never be in the same spin-orbital. This lecture shows how much the repulsion matters, where the Pauli principle comes from, how a Slater determinant builds it in, and why nobody stores the full many-electron wavefunction of a solid.
Press P to present one slide at a time, S for speaker notes.
How much does electron repulsion matter?
Ignore it and helium's energy is 30 eV wrong. A covalent bond is worth about 5 eV.
DFT & VASP Course · Lecture 52 / 16
Helium has two electrons and a nucleus of charge Z=2. Drop the repulsion between the electrons and the Hamiltonian splits into two hydrogen-like problems. Each electron sits in a 1s orbital with energy −Z2/2 hartree, so the total is −Z2=−4 hartree, or −108.8 eV.
Experiment says otherwise. The first electron takes 24.6 eV to remove and the second 54.4 eV, so the true ground-state energy is −79.0 eV. The non-interacting picture is off by 30 eV. For scale, breaking one O–H bond in water takes about 5.1 eV, and a typical strong covalent bond is worth about 5 eV: the error is six bonds' worth.
Adding the repulsion back as an average over the non-interacting wavefunction costs +85Z hartree =+34 eV and gives −74.8 eV, only 4 eV off. Those last 4 eV are still close to a whole covalent bond's worth, and closing that gap is what the rest of Part I is about.
Where 5Z/8 comes from
The average repulsion is the Coulomb integral J=∬n1(r1)n1(r2)/∣r1−r2∣dr1dr2 of two 1s charge clouds, each n1(r)=(Z3/π)e−2Zr. Do it in two steps. First, the electrostatic potential that one cloud creates at distance r from the nucleus, by Gauss's law (the charge inside r acts as a point charge, each shell outside contributes a constant):
v(r)=r1−(Z+r1)e−2Zr.Far away this is 1/r, the potential of one electron's charge; near the nucleus it is finite, Z, because the charge is spread out. Second, the energy of the other cloud in this potential, J=∫n1(r)v(r)4πr2dr, is an elementary integral of exponentials, and it gives J=5Z/8 hartree: for helium, 1.25 Ha = 34 eV.
What if electrons did not interact?
Then the Hamiltonian splits into one-electron pieces, and the wavefunction is a product of orbitals.
- Separable Hamiltonian: each electron feels only the nucleus; 1/∣r1−r2∣ is dropped.
- Product wavefunction: electron 1 in orbital a, electron 2 in orbital b, independently.
- Energies add: for helium, 2×(−2 Ha)=−4 Ha.
DFT & VASP Course · Lecture 53 / 16
First ignore the exchange-symmetry requirement so we can see separation clearly; the following slides impose it. Without the electron–electron term, the Hamiltonian is a sum of one-electron Hamiltonians, h^(1) acting only on electron 1 and h^(2) only on electron 2. Try a product Ψ=ϕa(r1)ϕb(r2), where h^ϕa=εaϕa and h^ϕb=εbϕb:
H^indΨ=[h^(1)ϕa(r1)]ϕb(r2)+ϕa(r1)[h^(2)ϕb(r2)]=(εa+εb)Ψ.So the total wavefunction is a product of orbitals and the total energy a sum of orbital energies. This is the picture behind every orbital diagram you have drawn. In a product, the probability density factorizes, ∣Ψ∣2=∣ϕa(r1)∣2∣ϕb(r2)∣2: where electron 1 is tells you nothing about where electron 2 is.
The electron–electron repulsion 1/∣r1−r2∣ depends on both positions at once, so with it the Hamiltonian no longer separates, and the exact Ψ is not a product: where one electron is depends on where the other is. Even without repulsion, a physical electron state must be antisymmetrized. A bare product is a building block with a second problem, which has nothing to do with energy.
What is electron spin?
Spin is an intrinsic two-state degree of freedom; combining it with a spatial orbital gives a spin-orbital.
- Two valuesms=+21 (α, “up”) or ms=−21 (β, “down”).
- Spin-orbitalχ(x)=ϕ(r)σ(s): a space part times a spin part.
- Arguments of the wavefunctionxi=(ri,si): three spatial coordinates plus an internal spin label.
- One spatial orbital, two spin-orbitals.
DFT & VASP Course · Lecture 54 / 16
Electrons carry an intrinsic angular momentum, spin, whose component along any chosen axis takes only two values, ms=±21 (in units of ℏ). It shows up in experiment: in 1922 Stern and Gerlach sent a beam of silver atoms through an inhomogeneous magnetic field and saw it split in two, a splitting later understood as the spin of the atom's one unpaired electron.
Intrinsic means a quantum property, rather than literal rotation of a tiny charged ball. Up and down refer to measurement along a chosen axis; a general spin state can be a superposition of those basis states. The spin label is an extra wavefunction argument, not a fourth direction in real space.
To include spin in a wavefunction, give each electron an internal label s that takes two values, and write x=(r,s) for position plus spin. The two spin functions, α(s) for up and β(s) for down, are orthonormal: summing over the two values of s, ⟨α∣α⟩=⟨β∣β⟩=1 and ⟨α∣β⟩=0. A product of a spatial orbital and a spin function is a spin-orbital, and two spin-orbitals with different spins are orthogonal even if their spatial parts are the same.
The collinear examples here use a fixed up/down spin basis. More general spinors can have both components depending on position. The spin-independent Hamiltonian of Lecture 1 does not act on spin explicitly; spin enters only through the symmetry rule on the next slides. (Spin–orbit coupling, which does act on it, is the subject of Lecture 24.)
What does “identical” mean for electrons?
Electron labels have no physical identity; exchanging them preserves observables and gives the fermion minus sign.
- The sign: swapping twice must give back Ψ, so the factor squares to 1. Nature fixes it: + for bosons (photons), − for fermions (electrons).
DFT & VASP Course · Lecture 55 / 16
Two electrons cannot be told apart. There is no “electron 1” and “electron 2” in nature, only labels we attach for bookkeeping, and swapping the labels cannot change the outcome of any measurement. So every probability must be unchanged by the swap, which means the wavefunction can change at most by a factor of magnitude one.
The exchange-sign statement uses the permutation-symmetry rule for identical particles in ordinary three-dimensional quantum mechanics. Equal position probabilities alone do not prove antisymmetry. In a definite exchange-symmetry state the exchange eigenvalue η squares to 1; the spin–statistics connection selects −1 for electrons (fermions) and +1 for bosons. We take that connection as a physical postulate here.
Write P^12 for the operator that swaps the labels, P^12Ψ(x1,x2)=Ψ(x2,x1). Swapping twice restores the original, P^122=1, so the eigenvalues of P^12 are +1 and −1: an acceptable wavefunction is either symmetric or antisymmetric under the swap. The Hamiltonian treats all electrons alike, so it commutes with P^12, and a state keeps its symmetry forever. Which sign electrons take is not a choice we make; the next slide gives nature's answer.
Which sign do electrons take?
Electron wavefunctions are antisymmetric: the same-spin pair density vanishes when the positions coincide.
DFT & VASP Course · Lecture 56 / 16
Particles with half-integer spin, fermions such as electrons, protons and neutrons, have antisymmetric wavefunctions; particles with integer spin, bosons such as photons, have symmetric ones. This is the spin–statistics theorem. Its proof belongs to relativistic quantum field theory, so in this course we take it as a postulate.
The simple product of spin-orbitals fails the test, because swapping the labels gives a different function. The fix is to subtract the swapped product:
Ψ=21[χa(x1)χb(x2)−χb(x1)χa(x2)].Swapping 1 and 2 turns the bracket into its negative, as required. When the two spin-orbitals have the same spin, the spin factor α(s1)α(s2) comes out, and the spatial part is [ϕa(r1)ϕb(r2)−ϕb(r1)ϕa(r2)]/2. That is what the figure shows, with ϕa and ϕb the box states ψ1 and ψ2.
On the diagonal x1=x2 the two terms cancel, so the antisymmetric ∣Ψ∣2 vanishes there: two electrons of the same spin are never found at the same place. The probability of finding them within a tenth of the box of each other is 0.016, against 0.200 for the plain product, more than ten times smaller. That region of reduced probability around each electron is the exchange hole. The symmetric combination does the opposite and piles the electrons up (0.385).
The figure's horizontal coordinate is electron 1's position and its vertical coordinate electron 2's position in the same 1D box. This square is a joint-density map in two-particle configuration space, rather than a 2D room. The diagonal represents equal positions. The exclusion concerns the density at coincidence; the probability of any exact point of a continuous distribution is already zero.
For two electrons in a definite total-spin state with a spin-independent Hamiltonian, spatial and spin factors can be separated. A singlet has total spin S = 0 and one spin state; a triplet has S = 1 and three spin-projection states. Their spatial and spin factors must multiply to give an antisymmetric state. So either the spatial part is antisymmetric and the spin part symmetric (the three triplet states, which include both spins up), or the spatial part is symmetric and the spin part antisymmetric (the singlet, [α(1)β(2)−β(1)α(2)]/2). Helium's ground state is a singlet. In the independent-orbital approximation, both electrons occupy 1s with opposite spins and the spatial part is 1s(r1)1s(r2), the same product as before. That is why antisymmetry does not change helium's estimates on the first slide; exchange acts between electrons of the same spin, and its effects show up as soon as two electrons share one.
What does a 2×2 determinant actually do?
Multiply across each diagonal and subtract: swapping the electron rows changes the sign.
DFT & VASP Course · Lecture 57 / 16
Each column samples one spin-orbital at both electron coordinates. Each row samples all the chosen spin-orbitals at one electron coordinate. The 2×2 determinant is a signed sum of two assignments: χa(x1)χb(x2)−χb(x1)χa(x2).
Press Swap electron labels. The two rows exchange and the determinant changes sign. Its square stays the same, preserving probability density. Choose Same spin-orbital: identical columns make the two products equal, so the determinant vanishes. Opposite-spin states of the same spatial orbital are different spin-orbitals; they do not create identical columns.
The next slide generalizes this to N electrons. The factorial N!=N(N−1)⋯1 counts possible assignments: 2! = 2, 3! = 6. For orthonormal occupied spin-orbitals, 1/N! normalizes the signed sum.
How do you build an antisymmetric wavefunction?
Write the spin-orbitals as a determinant: swapping two electrons swaps two rows and flips the sign.
- Rows are electrons, columns are spin-orbitals. Swapping two electrons swaps two rows, which flips the sign of any determinant.
- Normalization for orthonormal spin-orbitals; for N=2, the 1/2 of the previous slide.
DFT & VASP Course · Lecture 58 / 16
For more than two electrons, writing out all the signed products by hand quickly becomes impossible: three electrons already need six terms, ten electrons over three million. In 1929 John Slater found a compact solution in the properties of determinants. Put the spin-orbitals in the columns and the electrons in the rows. For two electrons,
21χa(x1)χa(x2)χb(x1)χb(x2)=21[χa(x1)χb(x2)−χb(x1)χa(x2)],exactly the wavefunction of the previous slide. For N electrons, expanding the determinant produces all N! assignments of electrons to orbitals, each with the sign of its permutation. Two properties of determinants then do all the work: swapping two rows flips the sign, which is antisymmetry; and two equal columns give zero, which is the next slide.
A Slater determinant is the simplest antisymmetric many-electron wavefunction. The determinant specifies a set of occupied spin-orbitals, with no assignment of one identifiable electron to one orbital. It is the starting point of Hartree–Fock theory (Lecture 6) and of the Kohn–Sham construction (Lecture 9).
Where does the Pauli exclusion principle come from?
A determinant with two equal columns is zero: no two electrons can occupy the same spin-orbital.
- Two electrons in the same spin-orbital: the wavefunction vanishes everywhere.
- Pauli exclusionAt most one electron per spin-orbital, so at most two per spatial orbital, with opposite spins.
- ConsequenceOrbital capacities are limited. Energies and electron interactions determine filling order.
- In VASPAt most 2 electrons per band without spin polarization; at most 1 per spin channel with it.
DFT & VASP Course · Lecture 59 / 16
A determinant with two identical columns is zero. Two identical columns would mean two electrons in the same spin-orbital, so such a wavefunction does not exist: this is the Pauli exclusion principle. Pauli proposed it in 1925 as a rule to explain atomic spectra; antisymmetry shows that it is not a separate law but a consequence of electrons being identical fermions. Pauli limits capacity; it does not by itself rank orbital energies or prove the aufbau order. Energies and interactions decide which allowed filling is lowest.
Since spin-orbitals with different spins are different, each spatial orbital can hold two electrons, one with each spin. That is why VASP's OUTCAR shows occupations of 2 for the filled bands of Lab 0, and why a spin-polarized run (ISPIN = 2), which treats up and down electrons in separate channels, shows occupations of 1 in each channel.
Is one Slater determinant the exact answer?
No. It enforces antisymmetry, but electrons also avoid each other because they repel: correlation.
- Built in by a determinantExchange: same-spin electrons keep apart.
- Missing from one determinantCorrelation: all electrons, of either spin, avoiding each other because of 1/∣ri−rj∣.
- HeliumFirst estimate −74.8 eV, measured −79.0 eV. Letting the orbital relax (Lecture 6) gives −77.9 eV; the last 1.1 eV is correlation.
- Next lectureThe best single determinant (Hartree–Fock), and what it still misses: the correlation energy.
DFT & VASP Course · Lecture 510 / 16
A Slater determinant makes the wavefunction antisymmetric, and with it same-spin electrons avoid each other. But a determinant is still built from independent orbitals: for an electron of the other spin, where it is remains uncorrelated with where you are. Real electrons repel, so all of them avoid each other more than a single determinant allows. The missing piece is called correlation.
Some states cannot even be written as a single determinant. Helium's excited configuration 1s2s with one electron up and one down is an example: its singlet state is [1s(1)2s(2)+2s(1)1s(2)] times the singlet spin function, and expanding that product gives two determinants, not one. Lecture 6 finds the best single determinant, the Hartree–Fock wavefunction, and defines the correlation energy as what it leaves out: for helium, the last 1.1 eV.
Why does no code store the many-electron Ψ of a solid?
On a grid of g points per coordinate, Ψ needs g³ᴺ numbers; the density needs g³, whatever N is.
DFT & VASP Course · Lecture 511 / 16
Lecture 0 met the exponential wall as a count of parameters. Here is the same wall as storage. Put each coordinate on a grid of only g=10 points. One electron's orbital then takes 103 numbers. The many-electron wavefunction of the 8 valence electrons in Lab 0's two-atom silicon cell is one function of 24 coordinates: 1024 numbers, about 1.6×1025 bytes as complex numbers, before spin is even counted. That is more than all the data storage on Earth. The density of the same electrons is a function of 3 coordinates: 103 real numbers, 8 kB.
Ten points per coordinate is far too coarse for real accuracy, which only widens the gap. This count concerns a direct unrestricted many-electron grid, not a universal limit on wavefunction methods. Orbital expansions, coupled-cluster methods and other structured approximations avoid storing this grid and can treat substantially larger systems. Change N and the grid to see how fast the wall rises.
What do DFT codes store instead?
One-electron orbitals and the density: N functions of three coordinates, not one function of 3N.
- Exact ΨOne function of 3N coordinates (plus spins). Grows exponentially with N.
- Kohn–Sham orbitalsOne-electron orbitals (about N/2 occupied per k-point, two electrons each), each a function of 3 coordinates: WAVECAR.
- DensityOne function of 3 coordinates: CHGCAR.
DFT & VASP Course · Lecture 512 / 16
Kohn–Sham DFT (Lecture 9) never needs the interacting Ψ. It works with a fictitious system of non-interacting electrons whose wavefunction is a single Slater determinant of one-electron orbitals, constructed to have the same density as the real electrons. A determinant is fully specified by its N orbitals, so storing it costs N functions of three coordinates, N functions on a chosen spatial representation. The number of spatial coefficients can itself grow with system size, so this does not promise linear computational scaling. The orbitals, one per band at each k-point, are what VASP keeps in WAVECAR, and the density is what it writes to CHGCAR.
The two effects of this lecture, exchange and correlation, do not disappear. In Kohn–Sham DFT their energies are collected, together with a kinetic-energy correction, into the exchange–correlation energy of Lecture 0, the one piece that must be approximated. That is why its name combines them.
Lab 5: does silicon want to be magnetic?
Rerun Lab 0 with ISPIN = 2. Predict: no magnetization, the same energy, occupations of 1.
INCAR: add one line to Lab 0
ISPIN = 2 # spin-polarized # MAGMOM defaults to 1.0 per atom
Same POSCAR, KPOINTS and POTCAR, in a fresh directory (no WAVECAR or CHGCAR from Lab 0).
Predict, then check
- OSZICAR, last line:
mag=is near 0, although MAGMOM started the run at 2 μB. - E0 approaches Lab 0's value as both calculations are converged.
- OUTCAR: two spin components, filled bands with occupation 1.
DFT & VASP Course · Lecture 513 / 16
ISPIN = 2 treats spin-up and spin-down electrons separately, so each spin channel can develop its own density and occupations. For a nonmagnetic system it is unnecessary and roughly doubles the cost. Without a MAGMOM line, each atom starts with a moment of 1.0 μB, so the run starts magnetic, 2 μB for the cell.
A magnetic moment measures magnetic response. Spin moments are reported in Bohr magnetons, μB, with μB=eℏ/(2me). In this collinear calculation without spin–orbit coupling, the total spin moment corresponds to the up-minus-down electron count in these units. MAGMOM supplies a starting guess, rather than fixing the final moment. Zero total moment can also result from equal and opposite local moments; it does not alone prove that every atom is nonmagnetic.
Silicon's electrons are all paired in filled bands, so the prediction is that the magnetization dies away during the SCF loop and the energy returns to Lab 0's. In a spin-polarized run, OSZICAR adds the total magnetization, mag=, to each ionic line. If your magnetization does not go to zero, that is worth bringing to the next session.
Which misconceptions should you drop now?
Five beliefs about many electrons. Click a card to turn it over.
“The Pauli principle is a repulsive force.”
It is a symmetry requirement on Ψ. No force term in the Hamiltonian produces it.
“Antisymmetry is just bookkeeping.”
It changes where electrons can be: same-spin electrons never meet, and orbital occupancy is restricted.
“A Slater determinant is the exact many-electron wavefunction.”
It includes exchange but misses correlation; some states need several determinants.
“Two electrons per orbital is a rule of thumb.”
It follows from antisymmetry: one per spin-orbital, two spins per spatial orbital.
“VASP stores the many-electron wavefunction.”
It stores one-electron orbitals (WAVECAR) and the density (CHGCAR).
DFT & VASP Course · Lecture 514 / 16
Check yourself
One question per outcome. Answer aloud, then click to compare.
Why does dropping electron repulsion make the problem easy, and how wrong is it for helium?
The Hamiltonian separates into one-electron pieces. For helium it gives −108.8 eV instead of −79.0 eV: 30 eV off.
What is a spin-orbital?
A spatial orbital times a spin function, α or β: χ(x) = φ(r)σ(s).
Why is a Slater determinant antisymmetric?
Swapping two electrons swaps two rows, and swapping two rows of a determinant flips its sign.
Why can't two electrons share a spin-orbital?
The determinant would have two equal columns and vanish: no such state exists.
Why does VASP store orbitals rather than Ψ?
Ψ of N electrons needs about g³ᴺ numbers on a grid; about N/2 doubly occupied orbitals need (N/2) × g³.
DFT & VASP Course · Lecture 515 / 16
What should you remember?
Electrons repel and are indistinguishable fermions: that is what makes many of them hard.
- Without repulsion the problem separates; for helium that is 30 eV wrong, six bonds' worth.
- Electrons have spin; a spin-orbital is a spatial orbital times α or β.
- Electron wavefunctions are antisymmetric; a Slater determinant builds that in.
- Antisymmetry gives Pauli exclusion and the exchange hole; it does not give correlation.
- Ψ of a solid cannot be stored; codes store one-electron orbitals and the density.
Next: Lecture 6 finds the best single determinant, Hartree–Fock, and measures what it misses.
DFT & VASP Course · Lecture 516 / 16
Key terms
- Electron spin
- An intrinsic angular momentum with two projection values ms=±21 along a chosen axis. Spin is an internal degree of freedom, not a fourth spatial dimension or literal spinning of a small ball.
- Spin-orbital
- A one-electron function of position and spin, χ(x)=ϕ(r)σ(s): a spatial orbital times a spin function, α (up) or β (down).
- Exchange hole
- The region around an electron where another electron of the same spin is unlikely to be found, because the antisymmetric wavefunction vanishes when two same-spin electrons meet.
- Slater determinant
- An antisymmetric many-electron wavefunction built as the determinant of N spin-orbitals, with electrons in the rows and orbitals in the columns. Swapping two electrons swaps two rows and flips the sign.
- Pauli exclusion principle
- No two electrons can occupy the same spin-orbital, so a spatial orbital holds at most two electrons, of opposite spin. It follows from the antisymmetry of the wavefunction.
- Electron correlation
- The way electrons of either spin avoid each other because they repel, beyond what a single Slater determinant describes. The correlation energy is the difference between the exact and the Hartree–Fock energy.
VASP tags in this lecture
ISPIN- 1: no spin polarization (spin-up and spin-down electrons share one density). 2: spin-polarized collinear calculation, with separate spin-up and spin-down orbitals and densities.Default: 1 · VASP Wiki: ISPIN
MAGMOM- Initial magnetic moment of each atom, used when no magnetization density is present yet; it also lowers the symmetry if it breaks it.Default: 1.0 per atom for ISPIN = 2; 3 × 1.0 per atom for noncollinear runs · VASP Wiki: MAGMOM