Bose-Einstein Distribution

The Bose–Einstein distribution answers one question exactly: how many indistinguishable bosons, on average, occupy a single-particle quantum state of energy $\varepsilon$ at temperature $T$ and chemical potential $\mu$ ? The answer, $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)$ , carries a $-1$ in the denominator that is responsible for everything distinctive about bosonic matter: the pile-up of low-energy occupation, the divergence that seeds Bose–Einstein condensation, and the Planck law of thermal radiation. We take from the quantum-mechanics track exactly what makes bosons bosons, and no more. The article aq-qm-bosons-fermions owns the microscopic rule: because a boson's many-body state is symmetric under exchange, "any number" of bosons may share the same single-particle state — in the owner's words, this is "why lasers and Bose–Einstein condensates exist," in contrast to fermions, where "put two in the same state and the antisymmetric combination cancels itself to zero." That article even forward-points to this one: it notes that fed into thermal ensembles, the symmetric counting "become[s] the Bose–Einstein and Fermi–Dirac distributions (statistical-mechanics track)." This article is that home. We do not re-derive the symmetry; we take unlimited occupancy as given and compute its thermodynamic consequence. The cleanest route is the grand canonical ensemble for an ideal quantum gas (aq-sm-grand-canonical-ideal-quantum-gases), where each single-particle level is an independent open subsystem. Its grand partition function is a geometric series over occupation number, $\Xi_\varepsilon=\sum_{n\ge0}(z\,e^{-\beta\varepsilon})^n$ , and a single logarithmic derivative gives the mean fill. Two features are ours to prove: that the chemical potential is forced below the ground-level energy, $\mu<\varepsilon_0$ , as a mathematical necessity of convergence — the pole where condensation begins; and that bosons bunch, their occupation variance $\langle n\rangle(1+\langle n\rangle)$ being super-Poissonian, with the single-mode correlation $\langle n(n-1)\rangle/\langle n\rangle^2$ exactly equal to $2$ . In the dilute, hot limit the $-1$ becomes negligible and the distribution collapses to Maxwell–Boltzmann, so one formula spans deep quantum degeneracy and the classical ideal gas.

Key takeaways

The Bose–Einstein distribution gives the mean occupation of a single-particle level of energy $\varepsilon$ for identical bosons: $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)$ .
It follows from summing a geometric series over the unbounded occupation number in the grand canonical ensemble — the "any number in the same state" rule that aq-qm-bosons-fermions owns for bosons.
The chemical potential must satisfy $\mu<\varepsilon_0$ (the ground-level energy) so occupations stay positive and finite; $\mu\to\varepsilon_0$ is where the ground state overflows and Bose–Einstein condensation begins.
Bosons cluster: the occupation variance is super-Poissonian, $\sigma_n^2=\langle n\rangle(1+\langle n\rangle)$ , and a single thermal mode has $\langle n(n-1)\rangle/\langle n\rangle^2=2$ exactly.
With $\mu=0$ (non-conserved quanta) the distribution becomes $1/(e^{\beta\varepsilon}-1)$ , the Planck factor of blackbody radiation and Debye phonons.
Key idea: treat each level as an independent open subsystem with no occupancy limit; the geometric sum $\sum_n x^n=1/(1-x)$ with $x=e^{-\beta(\varepsilon-\mu)}$ delivers $\langle n\rangle=x/(1-x)=1/(e^{\beta(\varepsilon-\mu)}-1)$ , and its convergence bound $x<1$ is the physics of condensation.

The core insight

Bosons obey no exclusion, so a single-particle level's occupation runs over $0,1,2,\dots$ without limit. In the grand canonical ensemble the level's partition function is the geometric series $\Xi_\varepsilon=1/(1-x)$ with $x=e^{-\beta(\varepsilon-\mu)}$ , and one logarithmic derivative gives $\langle n\rangle=x/(1-x)=1/(e^{\beta(\varepsilon-\mu)}-1)$ . The series converges only for $x<1$ , i.e. $\mu<\varepsilon_0$ — a purely mathematical bound that becomes the physical ceiling on the chemical potential and the onset of condensation.

Where this lives today

The Bose–Einstein distribution is one of the most experimentally alive formulas in physics. It governs blackbody radiation, from the cosmic microwave background to the design of thermal cameras, and lattice heat capacity through the Debye phonon gas. Its divergence at $\mu\to\varepsilon_0$ is Bose–Einstein condensation, realized in 1995 in ultracold atomic vapors and now a standard laboratory tool for quantum simulation, precision atom interferometry, and the study of superfluidity and vortices. Its bunching statistics, $g_2(0)=2$ for a thermal mode, underlie Hanbury Brown–Twiss intensity interferometry — used to measure stellar diameters — and serve as the baseline against which the antibunching of single-photon sources and the perfect coherence of lasers are defined in modern quantum optics.

The physical picture

Picture each quantum level as a box that holds any number of identical, sociable particles. The grand-canonical "price" of adding one more boson to a level of energy $\varepsilon$ is the factor $x=e^{-\beta(\varepsilon-\mu)}$ . With no exclusion to stop the pile-up, the box's contents follow a geometric distribution, and the mean fill is $x/(1-x)$ . Low-energy levels, where $x$ is near one, get heavily populated — bosons "want" to share. The $-1$ in the denominator versus the fermionic $+1$ is the entire difference between clustering and avoidance. Now raise the chemical potential. It floods every level with more particles, but for a Bose gas you cannot raise $\mu$ past the ground-level energy $\varepsilon_0$ without the lowest box overflowing: the factor $x$ for that box would reach one, the geometric series would diverge, and the mean occupation would run to infinity. This hard ceiling — a mathematical fact about geometric series — is the seed of Bose–Einstein condensation, a uniquely bosonic instability with no fermionic analogue. For photons and phonons, whose number is not conserved, $\mu=0$ identically and the distribution becomes the Planck factor $1/(e^{\beta\varepsilon}-1)$ .

Figure

Bose-Einstein occupation diverges as ε approaches μ, while the fermion curve is capped at 1

The punchline: the Bose curve is unbounded near $\varepsilon=\mu$ — the geometric series is diverging — while the Fermi curve is pinned below one. That divergence is not an error; it is where a macroscopic condensate forms.

Figure

Bunching: a thermal Bose mode has super-Poissonian variance, g₂(0) equals 2

Bosonic occupation fluctuates more than a Poisson process: the excess $\langle n\rangle^2$ is bunching, and the second-order coherence of a single thermal mode is exactly $g_2(0)=2$ .

The core equations

Each single-particle level of a Bose gas is an independent open subsystem exchanging particles with a reservoir. Its grand partition function sums the unbounded occupation with the fugacity $z=e^{\beta\mu}$ ; the sum is geometric and converges only when its ratio is below one, which is where the bosonic constraint on $\mu$ is born.

\Xi_\varepsilon = \sum_{n=0}^{\infty}\left(z\,e^{-\beta\varepsilon}\right)^{n} = \frac{1}{1-z\,e^{-\beta\varepsilon}}, \qquad z\equiv e^{\beta\mu} \qquad (1)

where: $\Xi_\varepsilon$ — grand partition function of one single-particle level; $z$ — fugacity; $n$ — occupation number, running over $0,1,2,\dots$ without limit (the bosonic "any number" rule). Physical reading: the box's contents are geometrically distributed; the closer $z\,e^{-\beta\varepsilon}$ is to one, the fatter the tail toward large occupation. Differentiating the log of $\Xi_\varepsilon$ with respect to $\mu$ (equivalently $z\,\partial_z$ ) pulls a factor $n$ into the sum, giving the mean occupation directly.

\langle n_\varepsilon\rangle = \frac{1}{\beta}\frac{\partial\ln\Xi_\varepsilon}{\partial\mu} = \frac{z\,e^{-\beta\varepsilon}}{1-z\,e^{-\beta\varepsilon}} = \frac{1}{e^{\beta(\varepsilon-\mu)}-1} \qquad (2)

where: $\langle n_\varepsilon\rangle$ — mean number of bosons in a level of energy $\varepsilon$ ; $\beta=1/k_BT$ ; $\mu$ — chemical potential (must satisfy $\mu<\varepsilon_0$ ). Physical reading: this is the Bose–Einstein distribution — the $-1$ enhances occupation of low levels and diverges as $\varepsilon\to\mu$ . The convergence of the geometric series is not a technicality; it is the physical constraint that defines a Bose gas. It bounds the chemical potential below the lowest level.

\mu < \varepsilon_0 \qquad (3)

where: $\varepsilon_0$ — the lowest single-particle energy. Physical reading: if $\mu$ reached $\varepsilon_0$ the ground-level occupation would diverge and a macroscopic condensate would form; $\mu$ can approach $\varepsilon_0$ but never cross it. The fluctuations expose the sociability of bosons. A second logarithmic derivative gives a super-Poissonian variance, larger than the Poisson value $\langle n\rangle$ by exactly $\langle n\rangle^2$ .

\sigma_n^2 = \langle n^2\rangle - \langle n\rangle^2 = \langle n\rangle\,(1+\langle n\rangle) \qquad (4)

where: $\sigma_n^2$ — variance of the occupation number; the excess $\langle n\rangle^2$ over the Poisson value is the bosonic bunching term. Physical reading: bosons arrive in correlated bunches; a single thermal mode has second-order coherence $g_2(0)=\langle n(n-1)\rangle/\langle n\rangle^2=2$ . Summed over the spectrum with a density of states or degeneracy weight, the distribution gives the total particle number, and in three dimensions the excited-state count is a polylogarithm that saturates — the arithmetic origin of condensation.

N - N_0 = \frac{V}{\lambda^3}\,\mathrm{Li}_{3/2}(z), \qquad \mathrm{Li}_{3/2}(1) = \zeta\!\left(\tfrac{3}{2}\right)\approx 2.612 \qquad (5)

where: $N_0$ — occupation of the ground state (macroscopic below $T_c$ ); $\lambda=h/\sqrt{2\pi m k_BT}$ — thermal de Broglie wavelength (aq-sm-thermal-de-broglie-wavelength-sm4); $\mathrm{Li}_{3/2}$ — polylogarithm; $V$ — volume. Physical reading: as $z\to1$ the excited states can hold at most $V\zeta(3/2)/\lambda^3$ particles; any surplus is forced into the ground state — the condensate, worked out in aq-sm-bose-einstein-condensation-ideal-gas.

Derivations

Derivation ABose-Einstein distribution from the single-level geometric sum

The goal is to derive $\langle n_\varepsilon\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)$ by summing the grand-canonical geometric series over an unbounded occupation, taking as our sole quantum input the rule owned by aq-qm-bosons-fermions that bosons place "any number in the same state." We work with the fugacity $z=e^{\beta\mu}$ so the algebra is transparent, and we make explicit where the bosonic constraint on $\mu$ enters — namely as the convergence condition of the very series we sum.

Step 1. Write the single-level grand weight. A level of energy $\varepsilon$ holding $n$ bosons has energy $n\varepsilon$ and particle number $n$ , so its grand-canonical weight is $(z\,e^{-\beta\varepsilon})^n$ . Because bosons in different levels are independent in the ideal gas, the full $\Xi$ factorizes and we focus on one level. Why: additivity of energy and number over levels makes each level a self-contained open subsystem (aq-sm-grand-canonical-ideal-quantum-gases).

Step 2. Sum the geometric series. With no exclusion the occupation runs to infinity, so $\Xi_\varepsilon=\sum_{n=0}^{\infty}(z\,e^{-\beta\varepsilon})^n=\dfrac{1}{1-z\,e^{-\beta\varepsilon}}$ , valid only if $z\,e^{-\beta\varepsilon}<1$ . Why: this is the exact place the "any number" rule and the $\mu<\varepsilon$ constraint both enter — an unbounded sum that converges only below the wall.

Step 3. Extract the mean occupation. Apply $z\,\partial_z\ln\Xi_\varepsilon$ , which brings a factor $n$ down inside the sum: $\langle n_\varepsilon\rangle=z\,\dfrac{\partial\ln\Xi_\varepsilon}{\partial z}=\dfrac{z\,e^{-\beta\varepsilon}}{1-z\,e^{-\beta\varepsilon}}$ . Why: the logarithmic derivative is the generating-function identity for the first moment of the geometric distribution.

Step 4. Rewrite in standard form. Substitute $z=e^{\beta\mu}$ and multiply numerator and denominator by $e^{\beta\varepsilon}/z$ : $\langle n_\varepsilon\rangle=\dfrac{1}{z^{-1}e^{\beta\varepsilon}-1}=\dfrac{1}{e^{\beta(\varepsilon-\mu)}-1}$ — the Bose–Einstein distribution. Why: the single algebraic rearrangement exposes the characteristic $-1$ and the dependence on $\varepsilon-\mu$ only. Verified: the geometric sum $\Xi_\varepsilon=1/(1-x)$ , the direct-moment identity $\langle n\rangle=(1-x)\sum_n n x^n=x/(1-x)$ , and the standard-form rewrite to $1/(e^{\beta(\varepsilon-\mu)}-1)$ are all checked symbolically in verify_bose-einstein-distribution.py.

Go deeper · Derivation B · 1 minThe convergence bound is the chemical-potential ceiling, and bosons bunch

Derivation BThe convergence bound is the chemical-potential ceiling, and bosons bunch

The goal is to prove the two structural results this article owns beyond the base menu: first, that $\mu<\varepsilon_0$ is not an assumption but the mathematical convergence condition of the geometric series, with the ground-level occupation diverging exactly at the bound; second, that the single-mode occupation is super-Poissonian with second-order coherence $g_2(0)=2$ exactly — the precise statement of bosonic bunching. Both follow from the same geometric distribution $p_n=(1-x)x^n$ with $x=e^{-\beta(\varepsilon-\mu)}$ .

Step 1. Pin the chemical potential below the ground level. The series $\sum_n x^n$ converges only for $x<1$ ; applied to every level, the binding case is the lowest energy $\varepsilon_0$ , giving $z\,e^{-\beta\varepsilon_0}<1$ , i.e. $\mu<\varepsilon_0$ . Why: a geometric series with ratio $\ge1$ diverges, so a Bose gas simply cannot have $\mu\ge\varepsilon_0$ .

Step 2. Show the bound is a genuine divergence. At $\varepsilon_0=0$ and $\beta=1$ , the ground occupation $\langle n_0\rangle=1/(e^{-\mu}-1)$ climbs monotonically as $\mu\to0^-$ : it passes through $\approx9.5,\,99.5,\,999.5$ at $\mu=-0.1,-0.01,-0.001$ . Push $\mu$ above $0$ and the formula returns a negative number — the ensemble's way of saying it has left its domain. Why: the pole at $\mu=\varepsilon_0$ is the ground state demanding a macroscopic population, the onset of condensation.

Step 3. Compute the variance from the second cumulant. Using $\sigma_n^2=k_BT\,\partial_\mu\langle n\rangle$ and differentiating $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)$ gives $\sigma_n^2=\langle n\rangle(1+\langle n\rangle)$ — larger than the classical Poisson value $\langle n\rangle$ by $\langle n\rangle^2$ . Why: the fluctuation–response identity connects the width of the occupation distribution to how fast the mean responds to $\mu$ .

Step 4. Extract the second-order coherence. For the geometric distribution $p_n=(1-x)x^n$ , the factorial moment is $\langle n(n-1)\rangle=2x^2/(1-x)^2=2\langle n\rangle^2$ , so $g_2(0)=\langle n(n-1)\rangle/\langle n\rangle^2=2$ exactly, independent of temperature or energy. Why: the value $2$ is the universal signature of a single thermal (chaotic) mode — the bunching measured directly in the Hanbury Brown–Twiss experiment. Verified: the $\mu<\varepsilon_0$ convergence branch and the monotone divergence of $\langle n_0\rangle$ , the negative-occupation flag for $\mu>\varepsilon_0$ , the super-Poissonian variance $\langle n\rangle(1+\langle n\rangle)$ , and the exact $g_2(0)=2$ are all checked in verify_bose-einstein-distribution.py.

Limiting cases

The single formula $\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)$ contains several physically distinct regimes:

Regime	Result	Meaning
Classical (non-degenerate)	$e^{\beta(\varepsilon-\mu)}\gg1:\ \langle n\rangle\approx e^{-\beta(\varepsilon-\mu)}$	Tiny occupations: the $-1$ is negligible and Bose–Einstein collapses to Maxwell–Boltzmann (aq-sm-maxwell-boltzmann-classical-limit); valid when $n\lambda^3\ll1$ .
Near the ground state	$\mu\to\varepsilon_0^-:\ \langle n_0\rangle\to\infty$	Ground-level occupation diverges as $\mu$ approaches $\varepsilon_0$ ; the onset of condensation.
Photon / phonon gas	$\mu=0:\ \langle n\rangle=1/(e^{\beta\varepsilon}-1)$	Non-conserved quanta have zero chemical potential; the Planck blackbody law and Debye phonon statistics.
Low-energy / Rayleigh–Jeans	$\beta(\varepsilon-\mu)\ll1:\ \langle n\rangle\approx k_BT/(\varepsilon-\mu)$	Each low-energy mode carries $\sim k_BT$ ; for photons this is the Rayleigh–Jeans tail.

Common trapThe divergence is condensation, not a mistake

The chemical potential of an ideal Bose gas can never exceed the ground-state energy: $\mu<\varepsilon_0$ . Substituting $\mu>\varepsilon$ into the formula gives negative or infinite "occupations," which are unphysical — but the divergence as $\mu\to\varepsilon_0$ is not a calculational error; it is the physics of Bose–Einstein condensation, the ground state accepting a macroscopic population. A second trap: do not confuse the per-level occupation $\langle n\rangle$ with the number of particles in an energy shell. The latter is $\langle n\rangle$ times the density-of-states weight $g(\varepsilon)$ ; a level can have $\langle n\rangle\gg1$ while contributing modestly to $N$ if it is nearly alone at that energy — which is exactly why only the ground level condenses.

Worked examples

Example 1 — Thermal occupation of a photon mode, and its bunching.

Step 1. Setup. Take a photon mode of energy $\hbar\omega=0.025\,\mathrm{eV}$ (a mid-infrared thermal photon) at $T=300\,\mathrm{K}$ . Photons are not conserved, so $\mu=0$ and $\langle n\rangle=1/(e^{\beta\hbar\omega}-1)$ .

Step 2. Evaluate. With $k_BT=0.02585\,\mathrm{eV}$ , $\beta\hbar\omega=0.025/0.02585=0.9671$ , so $\langle n\rangle=1/(e^{0.9671}-1)=1/(2.630-1)=1/1.630=\mathbf{0.613}$ photons per mode — comparable to one, the regime where thermal radiation is neither purely classical nor deeply quantum.

Step 3. The fluctuations. The variance is $\sigma_n^2=\langle n\rangle(1+\langle n\rangle)=0.613\times1.613=\mathbf{0.989}$ , so $\sigma_n=0.994$ — nearly as large as the mean itself, far above the Poisson width $\sqrt{0.613}=0.783$ . The excess is bunching.

Step 4. Second-order coherence. Regardless of $\langle n\rangle$ , a single thermal mode has $g_2(0)=\langle n(n-1)\rangle/\langle n\rangle^2=\mathbf{2}$ . Thermal photons arrive twice as clumped as a coherent (laser) beam, for which $g_2(0)=1$ — the distinction Hanbury Brown and Twiss exploited.

Example 2 — The condensation temperature of a dilute rubidium gas.

Step 1. Setup. Take $^{87}$ Rb atoms (mass $m=1.443\times10^{-25}\,\mathrm{kg}$ ) at number density $n=2.6\times10^{19}\,\mathrm{m^{-3}}$ . Condensation sets in when the excited states saturate, $n\lambda^3=\zeta(3/2)=2.612$ .

Step 2. Find the wavelength at threshold. From $n\lambda^3=2.612$ , $\lambda=(2.612/n)^{1/3}=(2.612/2.6\times10^{19})^{1/3}=(1.005\times10^{-19})^{1/3}=\mathbf{4.65\times10^{-7}\ m}$ — about half a micron, enormous on an atomic scale.

Step 3. Convert to a temperature. Invert $\lambda=h/\sqrt{2\pi m k_BT_c}$ to get $T_c=h^2/(2\pi m k_B\lambda^2)$ . With $h=6.626\times10^{-34}$ , $T_c=(6.626\times10^{-34})^2/(2\pi\times1.443\times10^{-25}\times1.381\times10^{-23}\times(4.65\times10^{-7})^2)$ .

Step 4. Evaluate. Numerator $=4.39\times10^{-67}$ ; denominator $=2\pi\times1.443\times10^{-25}\times1.381\times10^{-23}\times2.16\times10^{-13}=2.70\times10^{-60}$ ; so $T_c=\mathbf{1.6\times10^{-7}\ K=162\ nK}$ — a nanokelvin, exactly the regime of the 1995 experiments. Condensation is a low-temperature phenomenon precisely because $\lambda$ must grow to the interparticle spacing.

Origins · 1 minbose-einstein statistics and how this was discovered

bose-einstein statistics

formulated by S. N. Bose and A. Einstein, 1924–25 · Bose–Einstein statistics began in 1924 when Satyendra Nath Bose, then in Dhaka, re-derived Planck's radiation law by treating photon states as indistinguishable — counting arrangements "as if permuted photons were one state." Einstein saw the depth of the idea, translated and published Bose's paper, and in 1924–25 extended the counting from photons to a gas of conserved atoms, where he predicted that below a critical temperature a macroscopic fraction would occupy the single lowest state. The distribution

\langle n\rangle=1/(e^{\beta(\varepsilon-\mu)}-1)

is the direct thermodynamic expression of that indistinguishable counting; the microscopic symmetry that permits unlimited occupancy is owned by aq-qm-bosons-fermions.

History

Origins · 1 minhow this was discovered

History

1900

— Planck introduces his radiation law, implicitly the Bose factor for photons

1924

— Bose derives Planck's law by counting photon states as indistinguishable

1924

— Einstein extends Bose's counting to a gas of material atoms

1925

— Einstein predicts that below a critical temperature bosons macroscopically occupy the ground state

1938

— London proposes Bose–Einstein condensation as the mechanism behind superfluid helium-4

1956

— Hanbury Brown and Twiss measure photon bunching, confirming super-Poissonian bosonic statistics

1995

— Bose–Einstein condensation achieved in dilute rubidium and sodium vapors at nanokelvin temperatures

Going further

Texts: Pathria & Beale, Statistical Mechanics (3rd ed.), Ch. 6–7; Kardar, Statistical Physics of Particles, Ch. 7; Reif, Fundamentals of Statistical and Thermal Physics, Ch. 9; Sethna, Statistical Mechanics, Ch. 7; Landau & Lifshitz, Statistical Physics Part 1, §54–56.
Microscopic owner: the "any number in the same state" rule for bosons is owned by aq-qm-bosons-fermions; this article develops its thermodynamic consequence and is the statistical-mechanics home that the QM article forward-points to.
Builds on: aq-sm-grand-canonical-ideal-quantum-gases (the factorization and general $\langle n\rangle$ ), aq-sm-occupation-number-representation, aq-sm-symmetrization-postulate-wavefunction-symmetry, aq-sm-indistinguishability-identical-particles.
Leads to: aq-sm-bose-einstein-condensation-ideal-gas (the $\mu\to\varepsilon_0$ divergence), aq-sm-fermi-dirac-distribution (the sign-flipped counterpart), aq-sm-maxwell-boltzmann-classical-limit, aq-sm-photon-gas-planck-radiation (the $\mu=0$ Planck law).
Provenance: AllQFT `sm` export (Pathria & Beale), aq-sm-bose-einstein-distribution, ref STAT.QSTAT.BEDIS.1205. The distribution, the $\mu<\varepsilon_0$ bound, the bunching variance, $g_2(0)=2$ , $\zeta(3/2)\approx2.612$ , and the Rb $T_c=162\,\mathrm{nK}$ are python-verified (17 checks, all PASS).

Why is there a −1 in the denominator, and where does it come from?

It comes from summing the geometric series over an unbounded occupation. For bosons a level can hold any number of particles, so its partition function is $\sum_{n\ge0}x^n=1/(1-x)$ , and the mean fill is $x/(1-x)=1/(e^{\beta(\varepsilon-\mu)}-1)$ . The $-1$ is the fingerprint of the "any number" rule; fermions, capped at one, get a two-term sum $1+x$ and hence a $+1$ . The single sign is the entire difference between clustering and exclusion.

What is special about μ approaching ε₀?

It is where the ideal Bose gas condenses. The ground-level occupation $\langle n_0\rangle=1/(e^{\beta(\varepsilon_0-\mu)}-1)$ diverges as $\mu\to\varepsilon_0^-$ . Physically, the excited states can hold only a finite number of particles, $V\zeta(3/2)/\lambda^3$ ; when you try to add more (by lowering $T$ or raising $n$ ), $\mu$ climbs to $\varepsilon_0$ and the surplus pours into the single ground state, forming a macroscopic condensate. The divergence is the signal, not a breakdown.

How does this recover the classical gas?

When the gas is dilute and hot — $e^{\beta(\varepsilon-\mu)}\gg1$ , equivalently $n\lambda^3\ll1$ — the $-1$ is negligible next to the exponential, and $\langle n\rangle\approx e^{-\beta(\varepsilon-\mu)}$ , the Maxwell–Boltzmann factor with small fugacity $z=e^{\beta\mu}\ll1$ . In this regime the Bose–Einstein, Fermi–Dirac, and Boltzmann distributions all coincide because levels are so sparsely occupied that whether double occupancy is allowed or forbidden makes no difference.

What is bunching, and how is it measured?

Bunching is the tendency of bosons to arrive in correlated groups. Quantitatively, a single thermal mode has occupation variance $\langle n\rangle(1+\langle n\rangle)$ , larger than the Poisson value $\langle n\rangle$ , and second-order coherence $g_2(0)=2$ . Hanbury Brown and Twiss measured exactly this in 1956 by correlating the intensity of thermal light at two detectors: photons from a chaotic source cluster in time, giving a correlation twice the uncorrelated baseline. Lasers, being coherent, show $g_2(0)=1$ ; single-photon sources antibunch with $g_2(0)<1$ .

Why does the excited-state number saturate at ζ(3/2)?

Summing the distribution over a 3D free-particle spectrum gives the excited-state number $N_{\mathrm{ex}}=(V/\lambda^3)\,\mathrm{Li}_{3/2}(z)$ . The polylogarithm $\mathrm{Li}_{3/2}(z)$ is increasing in $z$ but bounded: its maximum, reached at $z=1$ , is $\zeta(3/2)\approx2.612$ . So the excited states can hold at most $V\zeta(3/2)/\lambda^3$ particles. Below the critical temperature this ceiling is exceeded by the fixed total $N$ , and the excess $N_0=N-N_{\mathrm{ex}}$ sits in the ground state — the mathematics of condensation.

Does this article prove Bose-Einstein condensation?

It proves the seed: that $\langle n_0\rangle$ diverges as $\mu\to\varepsilon_0$ (a mathematical necessity), and that the excited-state capacity saturates at $\zeta(3/2)$ . The full thermodynamics of the transition — the critical temperature $T_c$ , the condensate fraction $N_0/N=1-(T/T_c)^{3/2}$ , and the heat-capacity cusp — is developed in aq-sm-bose-einstein-condensation-ideal-gas and aq-sm-thermodynamics-bec-transition. This article establishes why the divergence must happen; those articles work out what it does.

Bose-Einstein Distribution

Where this lives today

The physical picture

The core equations

Derivations

Limiting cases

Worked examples

History

Going further

Before this

Where this leads