Indistinguishability of Identical Particles

In classical mechanics every particle carries a continuous trajectory, so even identical particles can in principle be tracked and labeled: we may say "particle 1 is here, particle 2 is there." Quantum mechanics destroys this picture. Once two identical particles have overlapping wave packets there is no observable, even in principle, that distinguishes a configuration from the one obtained by swapping them. The statement that identical particles are indistinguishable is therefore not an approximation but a structural constraint on the theory — and it rewrites how statistical mechanics counts. We take the microscopic content from the quantum-mechanics track, where aq-qm-indistinguishability-identical-particles establishes that "no measurement attaches a persistent label" to a particle of a given species, so permuting two particles cannot produce a new physical state. This article does not re-derive that principle; it asks the statistical-mechanics question that follows: if a permutation is not a physical operation, how many microstates does an N-particle gas actually have? The answer changes everything downstream. A state of a labeled system is a point in the product Hilbert space of N single-particle spaces; the physical Hilbert space is only the symmetric (bosons) or antisymmetric (fermions) subspace, which is drastically smaller. In the dilute, high-temperature limit where multiple occupancy is rare, that subspace is approximately $1/N!$ of the labeled space — exactly the Gibbs correction that rescues the extensivity of the entropy and resolves the Gibbs paradox. When the gas is degenerate — the thermal de Broglie wavelength approaching the interparticle spacing — the $N!$ shortcut fails, because multiple-occupancy configurations must be counted exactly, and one passes to occupation-number bookkeeping and the Bose–Einstein and Fermi–Dirac distributions. Indistinguishability is the single principle from which the whole architecture of quantum statistics descends.

Key takeaways

Identical quantum particles are fundamentally indistinguishable: no measurement tells which is which, so permuting two of them is not a physical operation — it maps a state to itself.
The quantum-mechanics owner aq-qm-indistinguishability-identical-particles supplies the microscopic principle; this article supplies the counting consequence for statistical mechanics.
The naive product over labeled particles overcounts by permutations; the Gibbs $1/N!$ factor removes the overcounting — but only in the dilute limit where no two particles share a state.
The $1/N!$ correction is an approximation to true quantum counting: exact only when $n\lambda^3 \ll 1$ ; between the exact boson and fermion counts it sits at their average for small systems.
It restores the extensivity of the entropy (Sackur–Tetrode with the correct $5/2$ ) and resolves the Gibbs paradox of mixing identical gases ( $\Delta S_{\text{mix}}=0$ ).
Key idea: permuting identical particles maps a state to itself (up to a sign), so microstate counting must divide out unphysical relabelings — a change of bookkeeping that precedes any dynamics.

The core insight

A permutation of identical particles is not a physical operation: it maps a state to itself (up to a sign). The physical Hilbert space is only the symmetric or antisymmetric subspace, and every microstate count must reflect this. The classical Gibbs $1/N!$ is the dilute-limit shadow of this exact quantum bookkeeping.

Where this lives today

Indistinguishability stopped being a paradox and became a laboratory dial. In ultracold-atom experiments the degeneracy parameter $n\lambda^3$ is tuned across unity by cooling: above it the gas is a classical Maxwell–Boltzmann cloud, below it a Bose–Einstein condensate or a degenerate Fermi gas, with the crossover imaged directly. Quantum-gas microscopes count atoms site by site and measure the sub- or super-Poissonian number fluctuations that indistinguishability imposes. The same counting underlies the Hong–Ou–Mandel effect in quantum optics, where two indistinguishable photons entering a beam splitter always leave together — a two-line consequence of bosonic symmetry that powers photonic quantum computing.

The physical picture

Picture two wave packets that start well separated and then overlap. While separated you could label them; once they overlap and re-separate, you cannot say which emergent packet "was" particle 1 — nature keeps no serial number. The amplitude for the direct process and the amplitude for the exchanged process refer to the same physical event, so they must be combined coherently (added for bosons, subtracted for fermions), not treated as two classical alternatives. The counting consequence is cleanest in a box-and-coins cartoon. Put $N$ particles into $g$ single-particle states. If the particles carried labels, the count would be $g^N$ . If they are truly identical, swapping two particles that sit in different states makes no new arrangement, so we overcount by $N!$ — but only as long as no two particles share a state. When a state holds two or more identical particles, some of those $N!$ permutations do nothing at all, so the overcounting is less than $N!$ . This is exactly why the simple Gibbs factor breaks down at high density and full quantum statistics takes over.

Figure

Two particles in two states: distinguishable, boson, fermion, and classical-corrected counts

The punchline figure: for a small system the Gibbs count $g^N/N!=2$ equals neither the boson count (3) nor the fermion count (1) — it is their average. The $1/N!$ is exact only in the dilute limit.

Figure

Ratio of quantum counts to the Gibbs count approaching one as the number of available states grows

Both quantum-to-Gibbs ratios converge to 1 as the states available per particle grow — the density regime $n\lambda^3\to 0$ where double occupancy is negligible and $1/N!$ becomes exact.

The core equations

The whole story starts with the exchange operator. Exchanging two particles twice returns the original configuration, which constrains its eigenvalues before any dynamics enters.

P_{ij}\,\psi(\ldots,\xi_i,\ldots,\xi_j,\ldots) = \psi(\ldots,\xi_j,\ldots,\xi_i,\ldots),\qquad P_{ij}^2=\mathbb{1} \qquad (1)

where: $P_{ij}$ — operator swapping the complete coordinates (space and spin) of particles $i$ and $j$ ; $\xi_i$ — full coordinate set of particle $i$ ; $\mathbb{1}$ — identity. Physical reading: swapping is an involution, so it can only rescale a physical state by a number whose square is one. Because no observable can distinguish identical particles, $P_{ij}$ commutes with the Hamiltonian and every observable, so physical states are its eigenstates with eigenvalue $\pm1$ .

[P_{ij},H]=0\quad\Longrightarrow\quad P_{ij}\psi=\pm\psi \qquad (2)

where: $H$ — Hamiltonian, symmetric under exchange; $+1$ — totally symmetric (bosons); $-1$ — totally antisymmetric (fermions), the only one-dimensional representations of the symmetric group. Physical reading: the exchange sign is a conserved label, fixed once and for all by the species. The statistical payoff is the counting rule. Treating particles as labeled and independent gives the product partition function; dividing by $N!$ removes the unphysical relabelings, but only when double occupancy is rare.

Z_N^{\text{labeled}}=Z_1^N \quad\longrightarrow\quad Z_N=\frac{Z_1^N}{N!}\quad (n\lambda^3\ll1) \qquad (3)

where: $Z_1=V/\lambda^3$ — single-particle partition function; $\lambda=h/\sqrt{2\pi m k_B T}$ — thermal de Broglie wavelength (its $\lambda=h/p$ origin is owned by aq-qm-de-broglie-matter-waves); $N!$ — the count of label permutations. Physical reading: the $1/N!$ is the dilute-limit shadow of the exact symmetric/antisymmetric subspace dimension. The correction earns its keep by making the entropy extensive, which the labeled count fails to do.

S = N k_B\!\left[\ln\!\left(\frac{V}{N\lambda^3}\right)+\frac{5}{2}\right] \qquad (4)

where: $V/N$ — volume per particle (the $N$ in the log comes from $\ln N!\approx N\ln N-N$ ); $\tfrac52$ — constant from the momentum integral plus Stirling. Physical reading: $S$ depends only on $V/N$ and $T$ , so doubling the system doubles $S$ — extensivity, restored by indistinguishability.

Derivations

Derivation ATwo-particle enumeration: why 1/N! is only a dilute-limit approximation

The goal here is to make quantitative the claim that the Gibbs $1/N!$ correction is not exact but an approximation valid only when double occupancy is negligible. We do this by fully enumerating the smallest nontrivial system — two particles distributed over two single-particle states — under three counting rules (labeled, bosonic, fermionic) and comparing them against the naive $g^N/N!$ prescription. The comparison exposes precisely where and why the shortcut breaks, and it foreshadows the exact occupation-number counting used in the degenerate regime.

Step 1. Count the labeled configurations. With two labeled particles A and B and two states, each particle is placed independently, giving $g^N=2^2=4$ arrangements: both in state 1, both in state 2, A in 1 & B in 2, A in 2 & B in 1. Why this overcounts: the last two differ only by a relabeling that no experiment can detect.

Step 2. Count the true quantum configurations. For bosons the physical states are the symmetric combinations: both in state 1, both in state 2, and the single symmetric "one in each" — three states, the stars-and-bars count $\binom{g+N-1}{N}=\binom{3}{2}=3$ . For fermions antisymmetry forbids double occupancy, leaving only "one in each," so $\binom{g}{N}=\binom{2}{2}=1$ . Why the difference: the exchange sign keeps the symmetric doubles for bosons and cancels them for fermions.

Step 3. Compare with the Gibbs prescription. The classical corrected count is $g^N/N!=4/2!=2$ . This matches neither quantum count; in fact it equals their average, $(3+1)/2=2$ . Why it fails here: the $1/N!$ assumes every configuration was relabeled $N!$ times, but the double-occupancy configurations were not, so dividing by $N!$ undercounts them for bosons and phantom-counts a forbidden one for fermions.

Step 4. Recover exactness in the dilute limit. Grow the number of available states $g$ at fixed $N$ . The ratios boson $/(g^N/N!)$ and fermion $/(g^N/N!)$ both approach 1: at $g=2,4,10,100$ the boson ratio is $1.5,1.25,1.10,1.01$ and the fermion ratio is $0.5,0.75,0.90,0.99$ . Why: as states become abundant the fraction of configurations with any double occupancy vanishes, so all three counts merge. This is the regime $n\lambda^3\ll1$ . Verified: the enumeration (labeled 4, boson 3, fermion 1, Gibbs 2) and the convergence of both ratios to 1 as $g\to\infty$ are machine-checked in verify_indistinguishability-identical-particles.py.

Go deeper · Derivation B · 1 minHow the Gibbs 1/N! restores extensivity and the Sackur–Tetrode 5/2

Derivation BHow the Gibbs 1/N! restores extensivity and the Sackur–Tetrode 5/2

The goal is to show that the single factor of $1/N!$ is exactly what converts a non-extensive labeled entropy into the extensive Sackur–Tetrode form, and that it supplies the $\ln N$ that turns the volume $V$ into the volume per particle $V/N$ . This is the thermodynamic justification for the counting change: without it the entropy of a classical ideal gas is not additive, and the Gibbs paradox appears. The route is a direct free-energy calculation with Stirling's approximation.

Step 1. Write the corrected free energy. With $Z_N=Z_1^N/N!$ and $Z_1=V/\lambda^3$ , we have $F=-k_BT\ln(Z_1^N/N!) = -Nk_BT\ln Z_1 + k_BT\ln N!$ . Why the extra term matters: the $\ln N!$ is the entire difference from the labeled calculation, and it carries the extensivity fix.

Step 2. Apply Stirling. Using $\ln N!\approx N\ln N - N$ , the free energy becomes $F=-Nk_BT[\ln(V/\lambda^3) - \ln N + 1]=-Nk_BT[\ln(V/(N\lambda^3))+1]$ . Why: the $-\ln N$ moves inside the logarithm, converting $V$ to $V/N$ — an intensive argument.

Step 3. Differentiate for the entropy. With $\lambda\propto T^{-1/2}$ so $\lambda^3\propto T^{-3/2}$ , computing $S=-\partial F/\partial T$ yields $S=Nk_B[\ln(V/(N\lambda^3))+\tfrac52]$ . Why the 5/2: the $\tfrac32$ comes from differentiating the $T^{3/2}$ in $\lambda^{-3}$ and the extra $1$ from the Stirling term — together the Sackur–Tetrode constant.

Step 4. Confirm extensivity and the paradox resolution. Because $S$ depends only on $V/N$ and $T$ , scaling $(N,V)\to(2N,2V)$ doubles $S$ exactly; and removing a partition between two identical gases gives $\Delta S_{\text{mix}}=0$ , versus $2Nk_B\ln2$ for genuinely distinct gases. Why this is the whole point: the labeled count would give a spurious mixing entropy for identical gases, which is the Gibbs paradox. Verified: the symbolic derivative reproduces the $5/2$ constant, $S(2N,2V)=2S(N,V)$ , and $\Delta S_{\text{mix}}=0$ for identical gases in verify_indistinguishability-identical-particles.py.

Limiting cases

Regime	Condition	Counting behavior	Governing description
Classical / dilute	$n\lambda^3\ll1$	overcounting $\approx N!$ ; Gibbs exact	Maxwell–Boltzmann, $Z_N=Z_1^N/N!$
Onset of degeneracy	$n\lambda^3\sim1$	overcounting $<N!$ ; Gibbs fails	full quantum statistics required
Deeply degenerate	$n\lambda^3\gg1$	occupation numbers dominate	Bose–Einstein / Fermi–Dirac (SM5)
Two labeled particles	$g,N$ small	$g^N/N!$ = average of B, F	exact enumeration only
Classical distinguishable	different species	no $1/N!$ ; genuine mixing	$\Delta S_{\text{mix}}=2Nk_B\ln2$

Worked examples

Example 1 — Enumerate the microstates of two particles in two states.

Step 1. Labeled count. Two distinguishable particles A, B in two states give $2^2=4$ arrangements. Answer: 4.

Step 2. Boson count. Symmetric states are {both in 1}, {both in 2}, {one in each}, i.e. $\binom{2+2-1}{2}=\binom{3}{2}=\mathbf{3}$ .

Step 3. Fermion count. Antisymmetry forbids doubles, leaving only {one in each}, i.e. $\binom{2}{2}=\mathbf{1}$ .

Step 4. Gibbs prescription. $g^N/N!=4/2!=\mathbf{2}$ , which equals the average $(3+1)/2$ of the boson and fermion counts and matches neither — the promised breakdown at small system size.

Example 2 — Is argon at room conditions classical?

Step 1. Thermal wavelength. For argon ( $m=6.63\times10^{-26}$ kg) at $T=300$ K, $\lambda=h/\sqrt{2\pi m k_B T}$ . Numerically $\sqrt{2\pi m k_B T}=\sqrt{2\pi\cdot6.63\times10^{-26}\cdot1.381\times10^{-23}\cdot300}=4.15\times10^{-24}$ kg·m/s, so $\lambda=6.63\times10^{-34}/4.15\times10^{-24}=\mathbf{1.60\times10^{-11}\ m}$ .

Step 2. Number density. At 1 atm, $n=P/k_BT=101325/(1.381\times10^{-23}\cdot300)=2.45\times10^{25}\ \mathrm{m^{-3}}$ .

Step 3. Degeneracy parameter. $n\lambda^3=2.45\times10^{25}\cdot(1.60\times10^{-11})^3=\mathbf{9.9\times10^{-8}}$ .

Step 4. Verdict. Since $n\lambda^3\approx10^{-7}\ll1$ , argon at room conditions is deeply classical: the Gibbs $1/N!$ is accurate to parts in $10^7$ , and quantum statistics is irrelevant until the gas is cooled to microkelvin temperatures or compressed enormously.

Origins · 1 minSatyendra Nath Bose and how this was discovered

Satyendra Nath Bose

1894–1974 · In 1924 Bose recounted the photon states of blackbody radiation by treating permuted photons as one state — the first use of indistinguishable counting — and mailed the result to Einstein, who extended it to material particles. The counting change he introduced precedes any Hamiltonian and is the ancestor of the Gibbs factor and Bose–Einstein statistics.

History

1902

— Gibbs identifies the mixing paradox and the

1/N!

correction in Elementary Principles in Statistical Mechanics

1911

— Sackur and (1912) Tetrode fix the absolute entropy constant using the phase-cell size

h

1924

— Bose counts photon states as indistinguishable; Einstein extends to atoms (1924–25)

1926

— Fermi and Dirac independently construct the antisymmetric counterpart

1926

— Heisenberg and Dirac formulate the exchange symmetry of many-electron wavefunctions

1964

— Messiah and Greenberg give the general statement of the symmetrization postulate

Going further

Texts: Pathria & Beale, Statistical Mechanics (3rd ed.), §§1.5–1.6, Ch. 5; Kardar, Statistical Physics of Particles, Ch. 7; Sethna, Statistical Mechanics: Entropy, Order Parameters, and Complexity, §7.1.
Microscopic owner: the quantum-mechanics track owns the identity principle at aq-qm-indistinguishability-identical-particles and the $\lambda=h/p$ scale at aq-qm-de-broglie-matter-waves; this article owns the statistical counting consequence.
Leads to: Gibbs correction and correct Boltzmann counting, the occupation-number representation, bosons/fermions and the spin-statistics connection, and the grand-canonical treatment of ideal quantum gases.
Provenance: AllQFT `sm` export (Pathria & Beale), migrated as aq-sm-indistinguishability-identical-particles, ref STAT.QSTAT.INDIS.1199. Every counting and entropy claim python-verified (15 checks, all PASS).

Is indistinguishability an approximation or exact?

It is exact. Identical particles of one species are indistinguishable in principle: no measurement attaches a persistent label. What is approximate is the classical shortcut $1/N!$ for counting, which reproduces the exact symmetric/antisymmetric subspace dimension only when double occupancy is negligible ( $n\lambda^3\ll1$ ). So the principle is a law; the $1/N!$ is its leading-order consequence.

Why divide by N! and not something else?

Because a labeled count assigns each of $N$ particles independently, and every physical configuration with all particles in distinct states is thereby counted once for each of the $N!$ orderings of the labels. Dividing by $N!$ removes exactly those relabelings. The subtlety is configurations where two particles share a state: those were not counted $N!$ times, so the division is only correct when such configurations are rare.

Does the Gibbs 1/N! match the boson or the fermion count?

Neither, in general. For two particles in two states the Gibbs count is 2, the boson count is 3, and the fermion count is 1 — the Gibbs value is their average. Only as the number of available states per particle grows large do all three converge, which is why $1/N!$ works for ordinary gases but must be replaced by exact occupation counting in the degenerate regime.

What does the thermal de Broglie wavelength have to do with counting?

It sets the length scale over which wave packets overlap. When $\lambda$ is much smaller than the mean spacing $n^{-1/3}$ , particles rarely share a quantum state and the classical count is accurate. When $\lambda$ approaches the spacing ( $n\lambda^3\sim1$ ), overlaps are common, multiple occupancy matters, and the $1/N!$ fails. The scale $\lambda=h/\sqrt{2\pi m k_B T}$ traces to the de Broglie relation $\lambda=h/p$ owned by aq-qm-de-broglie-matter-waves.

How does this resolve the Gibbs paradox?

The paradox is that a labeled (distinguishable) count predicts a nonzero entropy of mixing even when two gases are identical. Indistinguishability, encoded as the $1/N!$ , makes the entropy depend only on $V/N$ , so mixing two identical gases changes nothing: $\Delta S_{\text{mix}}=0$ . For genuinely distinct gases the entropy does increase by $2Nk_B\ln2$ , which is physical and measurable.

Is bosonic or fermionic counting "more" than classical?

Bosonic counting exceeds the Gibbs value (more symmetric configurations survive, including doubles), while fermionic counting is less (exclusion removes configurations). The Gibbs count sits between them and equals their average for small systems. This bracketing is why real bosons show enhanced ("bunched") occupancy and fermions show suppressed ("antibunched") occupancy relative to a classical gas.

Do I ever need the exact count instead of 1/N!?

Yes — whenever the gas is degenerate. Liquid helium, electrons in metals, neutron-star matter, and ultracold atomic gases all have $n\lambda^3\gtrsim1$ , so double occupancy is common and the $1/N!$ is wrong. There one passes to the occupation-number representation and the grand canonical ensemble, producing the Bose–Einstein and Fermi–Dirac distributions (SM5). For everyday gases at room temperature the classical shortcut is accurate to parts in $10^7$ .