Bose-Einstein Statistics

Definition

Bose-Einstein statistics is the rule for how indistinguishable integer-spin particles — bosons — distribute themselves over the available energy states in thermal equilibrium. The single number you need is the mean occupation of a state of energy E:

n(E) = 1 / (e^((E − μ)/kT) − 1)

Here μ is the chemical potential, k = 1.381 × 10⁻²³ J/K is Boltzmann's constant, and T is temperature. Writing β = 1/kT and ΔE = E − μ, the compact form is:

n = 1 / (e^(β·ΔE) − 1)

The entire physics of the boson world lives in that lonely − 1 in the denominator. Fermions get a + 1 instead (Fermi-Dirac statistics) and are capped at one particle per state; the classical Maxwell-Boltzmann distribution has neither and just gives n = e^−(E−μ)/kT. The minus sign is what allows the occupation to grow without bound as E → μ, and it is the seed of every dramatic bosonic phenomenon: blackbody radiation, superfluidity, lasers, and Bose-Einstein condensation.

How it works

Imagine a single quantum state — say, one mode of a cavity — and ask how many particles sit in it. In the grand-canonical ensemble each occupied particle costs a Boltzmann factor e^−(E−μ)/kT. For bosons, the state can hold n = 0, 1, 2, 3, … particles with no upper limit, so the probability weights form a geometric series. Summing it:

Z₁ = Σ_{n=0}^∞ x^n = 1/(1 − x),   where x = e^−(E−μ)/kT
⟨n⟩ = x · dZ₁/dx / Z₁ = x/(1 − x) = 1/(e^((E−μ)/kT) − 1)

The geometric series converges only if x < 1, i.e. if E > μ. That convergence condition is not a technicality — it is the statement that the chemical potential must lie below every available energy level. For a gas of massive bosons that means μ < E₀, the ground-state energy. Push μ right up against E₀ and the ground-state occupation runs away to infinity. That is the trapdoor through which condensation happens.

Two regimes are worth holding in your head:

• High energy / dilute (E − μ ≫ kT): the exponential is huge, the −1 is irrelevant, and n ≈ e^−(E−μ)/kT — pure classical Boltzmann behavior. States are nearly empty.
• Low energy (E − μ → 0⁺): expand the exponential, e^(βΔE) − 1 ≈ βΔE, so n ≈ kT/(E − μ) → ∞. The occupation diverges. This is the curve you see blowing up at the left edge of the visualization.

A worked example with concrete numbers

Take a single state sitting an energy ΔE = E − μ above the chemical potential and ask how its occupation changes as we cool it. Plug numbers into n = 1/(e^(ΔE/kT) − 1). At room temperature kT ≈ 0.0259 eV (T = 300 K), so for a state ΔE = 0.0259 eV above μ, the exponent βΔE = 1.0:

βΔE = 1.0:   n = 1/(e¹ − 1)   = 1/(2.718 − 1) = 0.582
βΔE = 0.1:   n = 1/(e^0.1 − 1) = 1/0.1052      = 9.51
βΔE = 0.01:  n = 1/(e^0.01 − 1)= 1/0.01005     = 99.5
βΔE = 0.001: n = 1/(e^0.001 −1)= 1/0.0010005   = 999.5

Halving the gap-to-temperature ratio roughly doubles the occupation; pushing ΔE a thousand times closer to μ multiplies it a thousandfold. The divergence is the hyperbola n ≈ kT/ΔE. Compare a fermion in the same state: βΔE = 0.001 gives n = 1/(e^0.001 + 1) = 0.4998 — it can never exceed 1, no matter how cold. That single sign flip is the difference between an electron sea that stacks one-per-state and a boson gas that collapses into one state.

Now the condensation threshold. For an ideal 3D gas of bosons the critical temperature is set by the phase-space-density condition nλ³ = 2.612 (the Riemann zeta value ζ(3/2)), where λ = h/√(2πmkT) is the thermal de Broglie wavelength:

kT_c = (2πℏ²/m) · (n / 2.612)^(2/3)

For rubidium-87 (m ≈ 1.44 × 10⁻²⁵ kg) at a trapped density of ~2.5 × 10¹⁹ m⁻³, this gives T_c ≈ 170 nK — the value Cornell, Wieman, and the JILA team reached in 1995 to make the first dilute-gas condensate (2001 Nobel Prize). Below T_c the ground-state population fraction grows as N₀/N = 1 − (T/T_c)^(3/2): at T = 0.5 T_c already 65% of the atoms share the single lowest state.

Variants and regimes

Statistics	Particles	Occupation formula	Max per state	Spin	Canonical examples
Bose-Einstein	Bosons	1/(e^((E−μ)/kT) − 1)	Unlimited	Integer (0, 1, 2…)	⁴He, photons, phonons, Cooper pairs
Fermi-Dirac	Fermions	1/(e^((E−μ)/kT) + 1)	One (Pauli)	Half-integer (½, 3⁄2…)	electrons, protons, ³He
Maxwell-Boltzmann	Distinguishable / dilute	e^−(E−μ)/kT	Unlimited (n ≪ 1)	Any (classical)	hot dilute gases
Planck (photon gas)	Photons	1/(e^(ℏω/kT) − 1)	Unlimited	Spin 1, μ = 0	blackbody cavity radiation
Debye (phonon gas)	Phonons	1/(e^(ℏω/kT) − 1)	Unlimited	Lattice modes, μ = 0	solid heat capacity
Condensed phase	Bosons below T_c	μ → E₀; N₀/N = 1 − (T/T_c)^(3/2)	Macroscopic in ground state	Integer	BEC, superfluid ⁴He

The crucial distinction within Bose-Einstein statistics itself is whether the particle number is conserved. Atoms in a trap are conserved, so they carry a real chemical potential μ that you tune by cooling, and they condense. Photons and phonons are not conserved — the cavity walls or the lattice make and destroy them at will — so equilibrium fixes their number by demanding μ = 0. With μ = 0 the formula becomes the Planck/Debye occupation 1/(e^(ℏω/kT) − 1), and there is no condensation in the usual sense (though photon BEC in dye microcavities, 2010, engineers an effective μ).

Common pitfalls and misconceptions

• Thinking n is a probability. It is a mean occupation number, and for bosons it can be far larger than 1 — even millions. Fermi-Dirac's n is bounded by 1 and reads like a probability, but Bose-Einstein's does not.
• Letting μ exceed the ground state. If you ever compute a negative n you have set μ above some E. For massive bosons μ < E₀ always; it can only kiss E₀ from below, and that kiss is condensation.
• Assuming photons condense like atoms. Free photons have μ = 0 fixed by non-conservation, so their number simply falls as you cool (the cavity goes dark) rather than piling into a ground state. Atom number is fixed, so the surplus has nowhere to go but the ground state.
• Confusing degeneracy with statistics. A gas is "quantum degenerate" when nλ³ ≳ 1 — when wavefunctions overlap. That is when Bose-Einstein and Fermi-Dirac diverge from classical behavior. Above that, all three statistics agree.
• Ignoring state degeneracy g. The formula gives occupation per single-particle state. To count particles in an energy level with degeneracy g, multiply by g; to count a continuum, integrate n(E) against the density of states.
• Forgetting the symmetry origin. The −1 is not arbitrary — it follows from the symmetric many-body wavefunction of integer-spin particles via the spin-statistics theorem. Swap two bosons and the wavefunction is unchanged; that permissiveness is what allows shared states.

Applications

• Blackbody radiation. Setting μ = 0 and integrating 1/(e^(ℏω/kT) − 1) over photon modes reproduces the Planck spectrum and the T⁴ Stefan-Boltzmann law. Bose-Einstein statistics is literally why the cosmic microwave background has the shape it does.
• Heat capacity of solids. Debye treated lattice vibrations as a phonon gas obeying Bose-Einstein statistics with μ = 0, explaining the famous T³ low-temperature heat capacity that classical equipartition could never reproduce.
• Bose-Einstein condensates. Ultracold atom traps reach T_c ~ 100 nK, producing macroscopic matter waves used in atom interferometry, precision gravimetry, and analog models of cosmology and black holes.
• Superfluidity. Liquid helium-4 below 2.17 K flows with zero viscosity — a partial condensation of its bosonic atoms. Helium-3 (a fermion) only goes superfluid 1000× colder, after pairing into bosonic Cooper pairs.
• Lasers. Stimulated emission rate scales with (1 + n) for the target mode, so photons pile preferentially into an already-populated mode — bosonic enhancement is the engine of coherent light.
• Superconductivity. Electrons pair into bosonic Cooper pairs that condense into a single coherent state, carrying current without resistance.

Derivation and performance analysis

Start from the grand partition function for one single-particle state, summed over all occupations because there is no exclusion:

Ξ = Σ_{n=0}^∞ e^(−n(E−μ)/kT) = 1 / (1 − e^−(E−μ)/kT)   (E > μ required)
⟨n⟩ = kT · ∂(ln Ξ)/∂μ = 1 / (e^((E−μ)/kT) − 1)

To get the total particle number of a real gas you sum over all single-particle states. Splitting off the ground state explicitly is essential near T_c, because the integral over excited states saturates at a finite value while the ground-state term n₀ = 1/(e^((E₀−μ)/kT) − 1) can grow arbitrarily large:

N = n₀ + ∫ g(E) / (e^((E−μ)/kT) − 1) dE

The integral has a ceiling N_excited,max ∝ T^(3/2). Once N exceeds that ceiling — i.e. below T_c — the difference N − N_excited,max is forced into n₀, and a macroscopic fraction occupies a single state. This is a genuine phase transition with no interactions required; pure statistics drives it. Numerically the distribution is trivial to evaluate, just one exponential per state, but near μ → E its O(1/ΔE) divergence demands care: tabulate the ground state separately, and use the expansion n ≈ kT/(E − μ) − ½ when βΔE is small to avoid catastrophic cancellation in e^x − 1 (use expm1 in code). The sum nλ³ = ζ(3/2) ≈ 2.612 is the closed-form threshold that makes the transition temperature a one-line calculation rather than a numerical search.