Fermi-Dirac Statistics

Definition

Suppose you have a gas of identical fermions — electrons, say — in thermal equilibrium at temperature T. Fermi-Dirac statistics tell you the average number of those particles you'll find sitting in a single quantum state of energy E:

       1
n(E) = ─────────────────
       e^((E − μ)/kT) + 1

Here μ is the chemical potential (the energy cost of adding one more particle), k is Boltzmann's constant, and T is the absolute temperature. Writing β = 1/kT, the exponent is just β·ΔE where ΔE = E − μ, so the distribution is often quoted as n = 1/(e^(β·ΔE) + 1).

The whole story lives in that lonely +1 in the denominator. It is the algebraic fingerprint of the Pauli exclusion principle. Because it is there, n can never exceed 1: a state is either empty, half-likely, or — at most — certainly occupied by a single fermion. There is no piling up.

0 ≤ n(E) ≤ 1      ← the Pauli ceiling

How it works — building the Fermi sea

Imagine an empty ladder of energy levels and a bucket of electrons to pour in. A classical particle would let them all settle into the lowest rung. Fermions can't do that — exclusion forbids two identical electrons (same spin included) from sharing a state. So they stack: lowest rung, next rung, next, each level taking exactly the number of distinct states it offers.

Keep pouring until you run out of electrons. The level where the last electron lands is the Fermi energy E_F. Everything below it is full; everything above is empty. That solid block of occupied states is the Fermi sea, and its flat top is the Fermi level.

Now plug T = 0 into the distribution and watch the step appear. The exponent (E − μ)/kT blows up:

E < μ  →  exponent → −∞  →  e^(−∞) = 0  →  n = 1/(0+1) = 1   (full)
E > μ  →  exponent → +∞  →  e^(+∞) = ∞  →  n = 1/(∞+1) = 0   (empty)
E = μ  →  exponent = 0    →  e^0    = 1  →  n = 1/(1+1) = ½   (the surface)

That is a perfect step function — a vertical cliff at E = E_F = μ(0). The Fermi sea has a razor-sharp surface at absolute zero.

Heat the system and the cliff erodes. Electrons just below E_F can borrow ~kT of thermal energy and hop to empty states just above. The step rounds into an S-curve of width a few kT, centered exactly on μ — and note that n(μ) = 1/2 holds at every temperature, the one fixed point of the whole family of curves.

A worked example — copper at room temperature

Copper has one conduction electron per atom and a number density of about n = 8.5 × 10²⁸ electrons/m³. For a free-electron gas the Fermi energy is

E_F = (ħ²/2m)(3π²n)^(2/3)

Plugging in gives E_F ≈ 7.0 eV for copper. Convert that to a temperature — the Fermi temperature — by dividing by Boltzmann's constant:

T_F = E_F / k = (7.0 eV)/(8.617×10⁻⁵ eV/K) ≈ 81,000 K

Room temperature is 300 K, so the thermal energy kT ≈ 0.026 eV. Compare the two scales:

kT / E_F = 0.026 / 7.0 ≈ 0.0037  ≈  1/270

The smearing width is only about a third of a percent of the depth of the sea. To a copper electron, "room temperature" is essentially absolute zero — the step is still nearly vertical. Only the topmost ~0.4% of electrons are thermally active; the rest are locked in the sea with no empty seats to move into. This is exactly why a metal's electronic heat capacity is tiny and grows linearly with T:

C_electronic ≈ (π²/2) N k (T/T_F)

For copper at 300 K that factor (T/T_F) ≈ 1/270 suppresses the naive classical prediction (3/2)Nk by more than two orders of magnitude — the long-standing puzzle that Fermi-Dirac statistics finally solved in 1926.

Fermi-Dirac vs the other two statistics

There are exactly three ways to count occupation of quantum states, and they differ by a single character. Memorize this table and most of statistical mechanics falls out.

Property	Fermi-Dirac	Bose-Einstein	Maxwell-Boltzmann
Occupation formula	1/(e^(βΔE) + 1)	1/(e^(βΔE) − 1)	e^(−βΔE)
Particle type	Fermions (half-integer spin)	Bosons (integer spin)	Classical / distinguishable
Spin examples	e⁻, p, n, quarks	photon, phonon, ⁴He, Higgs	(idealized point particles)
Max occupation per state	1 (Pauli ceiling)	unlimited (can condense)	unlimited (but n ≪ 1)
Behavior at T → 0	sharp Fermi sea, n: 1 → 0 step	macroscopic ground-state pileup	not valid (quantum regime)
Sign of statistics term	+1	−1	0 (no ±1)
Dilute high-T limit	→ Maxwell-Boltzmann	→ Maxwell-Boltzmann	itself
Hallmark phenomenon	degeneracy pressure	BEC	ideal gas law

The key takeaway: when occupation is tiny (n ≪ 1) the ±1 is dwarfed by the exponential, and all three converge to the classical Maxwell-Boltzmann form. Quantum statistics only matter when states are crowded — that is, when the system is degenerate.

Variants and regimes

• The degenerate limit (T ≪ T_F). The step is nearly perfect. Metals, white dwarfs, and neutron stars all live here. Only a thin shell of width ~kT around E_F participates in thermal physics; the Sommerfeld expansion captures the small corrections in powers of (kT/E_F)².
• The classical limit (T ≫ T_F or low density). The chemical potential goes large and negative, occupation everywhere drops below 1, and Fermi-Dirac becomes Maxwell-Boltzmann. Electrons in a hot, dilute plasma behave classically.
• Semiconductors. Here the Fermi level sits in the forbidden band gap, where there are no states. Doping shifts μ toward the conduction band (n-type) or valence band (p-type). Because μ moves strongly with temperature and doping, the tail of the Fermi-Dirac function controls carrier concentration — the foundation of every transistor and diode.
• Astrophysical degeneracy. In a white dwarf the electrons are so compressed that E_F reaches ~0.5 MeV and degeneracy pressure — present even at T = 0 — supports the star against gravity up to the Chandrasekhar limit of ≈ 1.4 M☉. Push past that and you get a neutron star, held up by neutron degeneracy instead.
• Relativistic Fermi gas. When E_F approaches mc², the dispersion E = pc replaces E = p²/2m, which softens the equation of state and is what makes the Chandrasekhar mass a hard ceiling rather than an arbitrary one.

Common pitfalls and misconceptions

• Thinking n is a probability. For fermions n is both a probability and an average occupation, because the only allowed values are 0 and 1 — but for bosons that coincidence fails. Don't carry the intuition across; n is the mean occupation number of a state.
• Confusing Fermi energy with Fermi level. Strictly, E_F is the chemical potential at T = 0; "Fermi level" is the chemical potential at any T. Solid-state texts often use them interchangeably, which is fine for metals (μ ≈ E_F) but dangerous for semiconductors, where the Fermi level lives in the gap and moves a lot.
• Forgetting spin degeneracy. Each spatial level usually holds two electrons (spin up and spin down), not one. The Pauli ceiling is per quantum state, and spin is part of the state label.
• Assuming the sea is static. Even at T = 0 the filled states carry momentum — the Fermi sea is full of fast-moving electrons. The Fermi velocity in copper is ~1.6 × 10⁶ m/s, about 0.5% of light speed, with no thermal energy at all.
• Using Maxwell-Boltzmann for electrons in a metal. This was the classical blunder that predicted a huge electronic heat capacity that experiments never saw. The degeneracy factor (T/T_F) is the fix.
• Treating μ as constant. For a free-electron gas μ drifts down as μ(T) ≈ E_F[1 − (π²/12)(kT/E_F)²]. Negligible for metals; decisive for semiconductors.

JavaScript — the distribution and the Fermi sea

// Fermi-Dirac mean occupation. Energies in eV, T in kelvin.
const kB = 8.617e-5; // Boltzmann constant in eV/K

function fermiDirac(E, mu, T) {
  if (T === 0) return E < mu ? 1 : (E > mu ? 0 : 0.5); // sharp step at T=0
  const x = (E - mu) / (kB * T);
  // Numerically stable for large |x|
  return x > 40 ? Math.exp(-x) : 1 / (Math.exp(x) + 1);
}

// Copper: Fermi energy ≈ 7.0 eV
const E_F = 7.0;

console.log(fermiDirac(6.0, E_F, 0));     // 1   — deep in the sea
console.log(fermiDirac(7.0, E_F, 0));     // 0.5 — exactly at the surface
console.log(fermiDirac(8.0, E_F, 0));     // 0   — above the sea
console.log(fermiDirac(7.0, E_F, 300));   // 0.5 — still ½ at E = μ, any T
console.log(fermiDirac(7.1, E_F, 300));   // ~0.021 — kT ≈ 0.026 eV smears it

// Width of the smear: span from n = 0.9 down to n = 0.1 is ~4kT
function stepWidth(T) { return 4 * kB * T; } // in eV
console.log(stepWidth(300).toFixed(4));   // 0.1034 eV  (~1.5% of E_F)

// Fermi temperature: how "cold" room temperature really is for copper
const T_F = E_F / kB;
console.log(Math.round(T_F));             // 81234 K
console.log((300 / T_F).toExponential(2)); // 3.69e-3  — degeneracy factor

Where the formula comes from

Fermi-Dirac counting drops straight out of the grand canonical ensemble. Consider a single state of energy E that can hold either 0 or 1 fermion (Pauli). Its grand partition function sums over those two possibilities:

Ξ = Σ e^(−β(E − μ)·N)  over N = 0, 1
  = 1 + e^(−β(E − μ))

The mean occupation is the log-derivative with respect to μ:

⟨N⟩ = (1/β) ∂(ln Ξ)/∂μ
    = e^(−β(E − μ)) / (1 + e^(−β(E − μ)))
    = 1 / (e^(β(E − μ)) + 1)

That is the Fermi-Dirac distribution, derived from nothing but "0 or 1 particle allowed." Compare the boson case, where N runs 0, 1, 2, … to infinity; that geometric series gives the −1 of Bose-Einstein statistics instead of the +1. The single restriction — the Pauli ceiling — is the entire difference between a Fermi sea and a Bose condensate.

Discovered independently by Enrico Fermi and Paul Dirac in 1926, this distribution explained at one stroke why metals conduct, why their heat capacity is small, why white dwarfs don't collapse, and why the periodic table has its shell structure at all.

Where Fermi-Dirac statistics show up

• Metals and conduction. Only electrons within ~kT of E_F can scatter and carry current or heat. The Fermi surface — the boundary of the occupied sea in momentum space — controls conductivity, the Hall effect, and magnetoresistance.
• Semiconductors and electronics. The position of the Fermi level relative to the band edges sets carrier density. Doping moves it; that is how diodes, transistors, LEDs, and every chip on Earth work.
• White dwarfs and neutron stars. Degeneracy pressure, an outward push that exists even at T = 0, is what stops gravitational collapse below the Chandrasekhar limit (≈ 1.4 M☉).
• Heat capacity of solids. The linear-in-T electronic term, on top of the Debye phonon contribution, lets experimentalists separate electron and lattice physics by measuring C/T at low temperature.
• Stellar nucleosynthesis. Electron degeneracy in stellar cores governs the helium flash and the conditions for runaway carbon burning in Type Ia supernovae.
• Cold-atom and quantum-gas experiments. Laser-cooled fermionic atoms (like ⁶Li or ⁴⁰K) directly realize a tunable Fermi sea, letting physicists watch the step sharpen as they approach nanokelvin temperatures.