Gibbs Correction and Correct Boltzmann Counting

Write down the partition function of $N$ identical, non-interacting atoms by treating each one as a labelled subsystem and you get $Z_N = Z_1^{\,N}$ — clean, factorized, and quietly wrong. The free energy it produces is not extensive: double the gas and the entropy more than doubles, and sliding out the partition between two halves of the same gas appears to create entropy from nothing. That contradiction is the Gibbs paradox, and its cure is a single factorial: divide by $N!$ . This article owns why that factorial is exactly right, and where the argument comes from at the quantum level. The classical statement of the fix — "permuting identical particles is not a new microstate, so the labelled count overcounts by $N!$ " — is the subject of the classical-mechanics siblings aq-sm-gibbs-paradox and aq-sm-correct-boltzmann-counting, which own the paradox and the counting argument in the Reif/Pathria register. Here we take the quantum-statistics viewpoint: the $1/N!$ is not a classical postulate at all but the dilute limit of exact quantum state counting. When the phase-space density $n\lambda^3$ is tiny, essentially every occupied single-particle level holds at most one particle, and the exact bosonic and fermionic counts both collapse onto $Z_1^N/N!$ . The factorial is the fingerprint of quantum indistinguishability surviving into the classical regime. Getting the count right has real consequences and real limits. With the correction, the ideal-gas entropy becomes the Sackur–Tetrode formula, whose absolute value — containing Planck's constant through $\lambda$ — matches calorimetry for argon and the noble gases. The correction is invisible to pressure and internal energy (fixed- $N$ derivatives) but essential to the chemical potential, the entropy of mixing, and every equilibrium that compares particle numbers. And it is only valid while $n\lambda^3 \ll 1$ : once levels start to be multiply occupied, no single number $N!$ can fix the count, and the full Bose–Einstein or Fermi–Dirac statistics must take over.

Key takeaways

The naive labelled count $Z_N=Z_1^N$ makes the entropy non-extensive and manufactures a spurious $2Nk_B\ln 2$ on doubling — the Gibbs paradox.
Dividing by $N!$ (correct Boltzmann counting) removes the overcount from permuting identical particles among singly-occupied states, restoring extensivity.
The classical paradox and counting are owned by aq-sm-gibbs-paradox and aq-sm-correct-boltzmann-counting; this article derives the $1/N!$ as the dilute limit of exact quantum counting.
Enumerate two particles over $g$ states: the exact boson count is $\binom{g+1}{2}$ , the fermion count $\binom{g}{2}$ , and both approach the "classical" $g^2/2!$ as $g\to\infty$ — the factorial is where they meet.
The correction is unique: no volume-independent prefactor $c^N$ can restore extensivity; only $1/N!$ , through $-N\ln N + N$ , converts $V^N$ into $(V/N)^N$ .
$1/N!$ leaves $P$ and $E$ untouched but fixes $\mu$ , absolute $S$ , and mixing — and is exact only for $n\lambda^3\ll1$ .
Key idea: $1/N!$ is not an ad hoc classical patch but the exact quantum state count in the dilute limit, where double occupancy is negligible and the boson, fermion, and labelled-over- $N!$ counts all coincide.

The core insight

The Gibbs $1/N!$ is the dilute limit of exact quantum counting: when occupancies are almost always 0 or 1 ( $n\lambda^3\ll1$ ), the boson count $\binom{g+1}{N}$ , the fermion count $\binom{g}{N}$ , and the labelled count $g^N/N!$ all coincide. Dividing $Z_1^N$ by $N!$ restores the extensivity that a labelled classical count destroys.

Where this lives today

Correct Boltzmann counting is quietly load-bearing across physical chemistry and astrophysics. Every equilibrium constant, every dissociation and ionization balance, rests on a chemical potential $\mu=k_BT\ln(n\lambda^3)$ that only exists once the free energy is extensive — the Saha equation for stellar ionization and the law of mass action both inherit the $\lambda^3$ and the $1/N!$ directly. In modern atomic physics the breakdown of the correction is the headline: as a trapped alkali gas is cooled toward $n\lambda^3\sim1$ , the $N!$ counting fails and the gas crosses into Bose–Einstein condensation or Fermi degeneracy, the regime where the exact quantum statistics this article previews take over. The Sackur–Tetrode entropy remains a standard benchmark: its agreement with calorimetric absolute entropies is cited as one of the cleanest fingerprints of Planck's constant inside an otherwise classical formula.

The physical picture

Deal $N$ identical, unmarked chips into a large number $g$ of pigeonholes, at most one chip per hole because the holes are far more numerous than the chips. If the chips were labelled — chip A, chip B — then swapping A and B between two holes would be a new arrangement, and the labelled bookkeeping counts $g^N$ ways. But the chips are truly identical: swapping them changes nothing. For any set of $N$ occupied holes there are exactly $N!$ orderings of the labels that all describe the same physical configuration. Dividing by $N!$ deletes this fiction. The reason it matters is extensivity. Entropy is supposed to scale with system size: two identical flasks of gas side by side must carry $S = 2S_0$ , and pulling the partition between two halves of one gas must produce zero entropy. The labelled count violates both — it hands you an extra $Nk_B\ln 2$ every time you double the system. That surplus is not physics; it is the artifact of pretending identical atoms wear name tags.

Figure

The labelled count overcounts each configuration by N! identical permutations

The labelled count treats a swap of identical particles as a new state. For $N$ singly-occupied levels there are exactly $N!$ such fictitious copies, so the honest count divides by $N!$ .

The deeper story is that $N!$ is a memory of quantum mechanics. The exact quantum trace counts each many-body state once. In the dilute limit that trace is $Z_1^N/N!$ — the symmetrization simply throws away the $N!$ orderings of $N$ distinct momenta. So the factor we insert "by hand" into a classical formula is really the leading term of a quantum result, and it stops being exactly $N!$ precisely when quantum degeneracy sets in.

Figure

Exact quantum counts converge to the classical g-squared over 2 as states outnumber particles

The exact boson and fermion counts bracket the classical $g^2/2!$ and both approach it as states outnumber particles. Where they meet, $1/N!$ counting is exact — the dilute regime $n\lambda^3\ll1$ .

The core equations

The starting point is the factorized single-particle partition function; the whole issue is the prefactor in front of $Z_1^N$ . The single-particle piece for a free particle in a box carries the thermal wavelength, which is owned and derived in aq-sm-thermal-de-broglie-wavelength.

Z_1 = \frac{V}{\lambda^3}, \qquad \lambda = \frac{h}{\sqrt{2\pi m k_B T}} \qquad (1)

where: $Z_1$ — single-particle translational partition function; $V$ — volume; $\lambda$ — thermal de Broglie wavelength; $m$ — mass; $T$ — temperature; $h$ — Planck constant. Physical reading: $Z_1$ is the number of thermally accessible single-particle states, roughly the box volume measured in units of the quantum volume $\lambda^3$ . The corrected many-body partition function is $Z_1^N$ with the Gibbs factorial removing the label overcount. This is the object this article defends.

Z_N = \frac{1}{N!}\,Z_1^{\,N} = \frac{1}{N!}\left(\frac{V}{\lambda^3}\right)^{N} \qquad (2)

where: $N!$ — number of permutations of $N$ identical particles among distinct single-particle states. Physical reading: each physical configuration of singly-occupied states is reached by $N!$ labelled assignments, all counted once by the division. Feeding $Z_N$ through $F=-k_BT\ln Z_N$ with Stirling's approximation gives a manifestly extensive free energy per particle — the payoff of the correction.

\frac{F}{N} = -k_B T\left[\ln\!\left(\frac{V/N}{\lambda^3}\right)+1\right] \qquad (3)

where: $F$ — Helmholtz free energy; $V/N$ — volume per particle (intensive). Physical reading: after the correction the free energy depends only on the intensive density $n=N/V$ , so $F$ scales linearly with system size. The entropy that follows is the Sackur–Tetrode formula, whose absolute value is the experimental test of the whole construction.

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

where: $S$ — entropy; $V/N\lambda^3 = 1/(n\lambda^3)$ — inverse degeneracy parameter. Physical reading: the entropy per particle depends on how many quantum volumes each particle commands; the $5/2$ is $3/2$ from the $T$ -dependence of $\lambda^{-3}$ plus $1$ from $-F/T$ .

Derivations

Derivation AExtensivity restored: the N! cures the Gibbs paradox

The goal is to show, by direct computation, that the labelled count $Z_N=Z_1^N$ produces an entropy that fails the extensivity test — generating a spurious $2Nk_B\ln 2$ when the system is doubled — and that inserting $1/N!$ removes exactly that surplus, leaving the Sackur–Tetrode entropy which scales linearly with system size. This is the thermodynamic statement of the paradox; the concept is owned by aq-sm-gibbs-paradox, and here we make the additivity failure and its repair fully explicit.

Step 1. Start from the uncorrected free energy. With $Z_N^{\text{naive}}=Z_1^N=(V/\lambda^3)^N$ , we get $F^{\text{naive}}=-k_BT\ln Z_N^{\text{naive}}=-Nk_BT\ln(V/\lambda^3)$ . Why: $\ln Z_N^{\text{naive}}=N\ln(V/\lambda^3)$ with no $N$ -dependence beyond the overall factor $N$ .

Step 2. Differentiate for the entropy. Using $S=-(\partial F/\partial T)_{V,N}$ and $\lambda^{-3}\propto T^{3/2}$ so that $\partial_T\ln\lambda^{-3}=3/(2T)$ , one finds $S^{\text{naive}}=Nk_B[\ln(V/\lambda^3)+3/2]$ . Why: the temperature derivative of the $\ln\lambda^{-3}$ term contributes $Nk_B\cdot T\cdot 3/(2T)=\tfrac32 Nk_B$ .

Step 3. Test additivity. Replace $(N,V)\to(2N,2V)$ : $S^{\text{naive}}(2N,2V)=2Nk_B[\ln(2V/\lambda^3)+3/2]=2S^{\text{naive}}(N,V)+2Nk_B\ln 2$ . A surplus $2Nk_B\ln 2$ appears — doubling the system overproduces entropy. Why: the argument of the log is the extensive $V$ , not the intensive $V/N$ , so scaling $V$ leaks a $\ln 2$ per particle.

Step 4. Insert the Gibbs factor. With $Z_N=Z_1^N/N!$ and Stirling $\ln N!\approx N\ln N-N$ , the free energy becomes $F=-Nk_BT[\ln(V/N\lambda^3)+1]$ . The $-\ln N!$ supplies $-N\ln N+N$ , converting $V$ inside the log to $V/N$ . Why: $\ln(V/\lambda^3)-\ln N+1 = \ln(V/N\lambda^3)+1$ turns the extensive volume into an intensive density.

Step 5. Re-test additivity. Now $S(2N,2V)=2Nk_B[\ln(2V/2N\lambda^3)+5/2]=2Nk_B[\ln(V/N\lambda^3)+5/2]=2S(N,V)$ exactly — and removing a partition between identical gases at equal $T,p$ gives $\Delta S_{\text{mix}}=0$ . Why: $2V/2N=V/N$ is invariant under doubling, so the entropy is strictly linear in system size. Verified: the naive surplus $2Nk_B\ln 2$ , the Sackur–Tetrode $5/2$ from $S=-\partial F/\partial T$ , and the exact additivity $S(2N,2V)=2S(N,V)$ are checked symbolically in verify_gibbs-correction-boltzmann-counting.py.

Go deeper · Derivation B · 2 minThe 1/N! as the dilute limit of exact quantum counting

Derivation BThe 1/N! as the dilute limit of exact quantum counting

The goal is to demonstrate that $1/N!$ is not a classical postulate but the leading term of the exact quantum state count, by enumerating the true number of $N$ -particle states built from $g$ single-particle levels for bosons and fermions and showing both collapse onto the "classical" $g^N/N!$ when $g\gg N$ . This is the quantum-statistics content this article owns; the classical counting argument is owned by aq-sm-correct-boltzmann-counting, and here we ground it in exact combinatorics and the small-occupancy limit.

Step 1. Count exact quantum states for $N=2$ . From $g$ single-particle levels, the number of two-boson states (symmetric, any occupancy) is $\binom{g+1}{2}=g(g+1)/2$ , and the number of two-fermion states (antisymmetric, occupancy $0$ or $1$ ) is $\binom{g}{2}=g(g-1)/2$ . Why: bosons allow the two "same-level" states, fermions forbid them, differing by exactly the $g$ diagonal configurations.

Step 2. Compare to the classical count. The labelled count is $g^2$ , and dividing by $2!$ gives $g^2/2$ . Both quantum counts are $g^2/2 \pm g/2$ : bosons exceed the classical value by the double-occupancy states, fermions fall short by the excluded ones. Why: the $\pm g/2$ is precisely the same-level contribution, present for bosons, absent for fermions, and negligible against $g^2/2$ once $g\gg 2$ .

Step 3. Take the dilute limit. As $g\to\infty$ (equivalently $n\lambda^3\to 0$ , since $g\sim Z_1\sim V/\lambda^3$ ), the ratios $\binom{g+1}{2}/(g^2/2)=(g+1)/g\to 1$ and $\binom{g}{2}/(g^2/2)=(g-1)/g\to 1$ . Boson and fermion counts converge onto $g^N/N!$ . Why: double occupancy has vanishing probability when states vastly outnumber particles, so the statistics become irrelevant to the count.

Step 4. Read off the partition function. The many-body sum $\sum_{i_1\dots i_N}e^{-\beta\sum_k\varepsilon_{i_k}}=Z_1^N$ over-counts each singly-occupied configuration $N!$ times; dividing gives $Z_N\to Z_1^N/N!$ . The Gibbs factor is the exact quantum count in disguise. Why: summing each particle freely over all levels double-counts every ordering of $N$ distinct occupied levels, which is $N!$ .

Step 5. Establish uniqueness. Extensivity forces the prefactor: a volume-independent constant $c^N$ shifts $\ln Z_N$ by $N\ln c$ , which cannot convert $V$ to $V/N$ inside the log (that requires the $N$ -dependent $-N\ln N$ from $\ln N!$ ). Only $1/N!$ works. Why: restoring extensivity needs a term growing like $-N\ln N$ , which no constant-per-particle factor can supply. Verified: the exact counts $\binom{g+1}{2}$ , $\binom{g}{2}$ , their ratios $\to 1$ , and the uniqueness of $1/N!$ against a trial $c^N$ are checked in verify_gibbs-correction-boltzmann-counting.py.

Limiting cases

Regime / system	Prescription	Meaning
Dilute classical gas ( $n\lambda^3\ll1$ )	divide by $N!$	single occupancy dominates; Sackur–Tetrode holds — the domain of validity
Degenerate quantum gas ( $n\lambda^3\gtrsim1$ )	full Bose–Einstein / Fermi–Dirac	multiple occupancy matters; no single $N!$ can fix the count
Localized particles (crystal lattice)	do NOT divide by $N!$	atoms are labelled by fixed sites, hence distinguishable — Einstein solid
Mixing distinct species $A\ne B$	no shared $N_A!N_B!$ overcount	permuting an $A$ with a $B$ IS a new state; genuine $\Delta S_{\text{mix}}=-Nk_B\sum_i x_i\ln x_i$
Two particles, $g$ states	$\binom{g+1}{2}$ (B), $\binom{g}{2}$ (F)	exact counts; both $\to g^2/2!$ as $g\to\infty$

Worked examples

Example 1 — Enumerate the exact quantum counts and watch them meet the classical value.

Step 1. Take $g=3$ single-particle levels and $N=2$ identical particles. List the boson states (symmetric, any occupancy): $\{11\},\{22\},\{33\}$ (both in one level) and $\{12\},\{13\},\{23\}$ (one each). That is $\binom{3+1}{2}=\binom{4}{2}=\mathbf{6}$ states.

Step 2. List the fermion states (occupancy $\le 1$ ): only $\{12\},\{13\},\{23\}$ , i.e. $\binom{3}{2}=\mathbf{3}$ states — the three same-level configurations are excluded.

Step 3. The classical count is $g^2/2!=9/2=\mathbf{4.5}$ . Bosons ( $6$ ) sit above it, fermions ( $3$ ) below, differing from $4.5$ by $\pm 1.5=\pm g/2$ .

Step 4. Grow the box. At $g=1000$ : boson $\binom{1001}{2}=500500$ , fermion $\binom{1000}{2}=499500$ , classical $10^6/2=500000$ . The three now agree to $\pm 0.1\%$ . In a real gas $g\sim Z_1\sim 10^{30}$ , so the agreement is exact to $\sim10^{-30}$ — the regime where $1/N!$ is exact.

Example 2 — The Sackur–Tetrode entropy of argon, and the paradox surplus it removes.

Step 1. Take argon at $T=300$ K, $p=1$ atm, so $n=p/k_BT=1.013\times10^5/(1.381\times10^{-23}\times300)=2.45\times10^{25}\ \mathrm{m^{-3}}$ . With $m=6.63\times10^{-26}$ kg, $\lambda=h/\sqrt{2\pi m k_BT}=6.626\times10^{-34}/\sqrt{2\pi\cdot6.63\times10^{-26}\cdot1.381\times10^{-23}\cdot300}=1.60\times10^{-11}$ m.

Step 2. Form the degeneracy parameter $n\lambda^3=2.45\times10^{25}\times(1.60\times10^{-11})^3=2.45\times10^{25}\times4.10\times10^{-33}=\mathbf{1.0\times10^{-7}}$ — deeply classical, so $1/N!$ counting is valid.

Step 3. Sackur–Tetrode: $S/Nk_B=\ln(1/n\lambda^3)+5/2=\ln(1.0\times10^{7})+2.5=16.1+2.5=18.6$ . Per mole, $S_m=18.6\times R=18.6\times8.314=\mathbf{154.7\ J\,mol^{-1}K^{-1}}$ , matching the calorimetric argon value $154.8\ \mathrm{J\,mol^{-1}K^{-1}}$ .

Step 4. The surplus removed. Without $1/N!$ , doubling this sample would add a fake $2Nk_B\ln 2$ ; per mole of the doubled system that is $2R\ln 2=2\times8.314\times0.693=\mathbf{11.5\ J\,mol^{-1}K^{-1}}$ of entropy conjured from nothing. The factorial is what makes the absolute entropy come out right and the mixing of identical gases give zero.

Origins · 1 minquantum statistics and how this was discovered

quantum statistics

the dilute limit of quantum counting · The

1/N!

correction is best understood as the dilute limit of quantum statistics. Gibbs introduced the "generic phase" division in his 1902 Elementary Principles in Statistical Mechanics to rescue extensivity, a quarter-century before Bose (1924), Einstein (1924–25), Fermi and Dirac (1926) explained why permutations of identical particles must not be counted as distinct states at all. In the dilute, non-degenerate regime

n\lambda^3\ll1

the exact boson and fermion state counts both collapse to

g^N/N!

, so the classical

1/N!

is not an ad hoc fix but the leading term of the true quantum count — exact only when double occupancy is negligible.

History

1902

— Gibbs identifies the non-extensivity of the labelled ideal-gas entropy and the identical-gas mixing paradox

1911

— Sackur and (independently) Tetrode obtain the absolute entropy of a monatomic gas, fixing the constant with Planck's h

1912

— Tetrode's fuller treatment ties the entropy constant to the quantum of action

1924

— Bose's counting of photon states makes indistinguishable counting a physical postulate

1925

— Einstein extends the counting to a massive-atom gas, foreshadowing the classical N! as a quantum limit

1931

— Van Vleck and others clarify the classical limit of the quantum partition function, recovering Z₁ᴺ/N!

Going further

Texts: Pathria & Beale, Statistical Mechanics (3rd ed.), §1.5–1.6, §3.5 (Gibbs paradox, Sackur–Tetrode); Kardar, Statistical Physics of Particles, Ch. 4 (Gibbs correction) and Ch. 7 (classical limit of quantum gases); Reif, Fundamentals of Statistical and Thermal Physics, §7.3; Landau & Lifshitz, Statistical Physics Part 1, §31, §41.
Classical owners: the paradox itself is owned by aq-sm-gibbs-paradox; the classical counting argument by aq-sm-correct-boltzmann-counting. This article develops the quantum-statistics viewpoint (dilute limit of exact counting).
Builds on: Indistinguishability of Identical Particles · Thermal de Broglie Wavelength · Quantum Concentration and the Classicality Criterion.
Leads to: Maxwell–Boltzmann Classical Limit · Grand Canonical Ensemble for Ideal Quantum Gases · Occupation Number Representation.
Provenance: AllQFT `sm` export (Pathria & Beale), aq-sm-gibbs-correction-boltzmann-counting, ref STAT.QSTAT.GCBC.1208. Sackur–Tetrode $5/2$ , the $2Nk_B\ln 2$ surplus, $\mu=k_BT\ln(n\lambda^3)$ , the exact boson/fermion two-particle counts, and the uniqueness of $1/N!$ are all python-verified (all checks PASS).

Why exactly N! and not some other function of N?

Because in the dilute regime every occupied single-particle state holds exactly one particle, and a set of $N$ distinct occupied states can be reached by $N!$ different labelled assignments — one for each ordering of the labels. All $N!$ describe the same physical (unlabelled) configuration, so the honest count divides by $N!$ . If double occupancy were common, the overcount for a configuration with occupation numbers $\{n_s\}$ would be $N!/\prod_s n_s!$ rather than $N!$ , which is exactly why the simple factorial fails once $n\lambda^3\sim1$ .

Does the 1/N! change the pressure or energy?

No. The factor $N!$ contributes a term $-\ln N!$ to $\ln Z_N$ that depends only on $N$ , not on $V$ or $\beta$ . Pressure $P=k_BT(\partial\ln Z_N/\partial V)_{T,N}$ and energy $E=-(\partial\ln Z_N/\partial\beta)_{V,N}$ are derivatives at fixed $N$ , so the $N$ -only term drops out. What the correction does change are the chemical potential (an $N$ -derivative), the absolute entropy, and any free energy of mixing — the quantities that compare different particle numbers.

Is the Gibbs correction a classical idea or a quantum one?

Both, historically layered. Gibbs introduced the $1/N!$ in 1902 as a classical device to restore extensivity, before anyone knew why identical particles cannot be labelled. Quantum mechanics later supplied the reason: identical particles are genuinely indistinguishable, and the exact many-body state count already builds this in. In the dilute limit that exact count reduces to $Z_1^N/N!$ , so the classical patch turns out to be the leading term of the quantum result — which is the viewpoint this article develops.

Why must you NOT divide by N! for a crystal?

Because atoms sitting on distinct, fixed lattice sites are distinguishable by their positions. The atom at site $\mathbf r_1$ is a physically different subsystem from the one at $\mathbf r_2$ ; swapping them is not a relabelling of interchangeable particles but a move between labelled sites, and it does not produce an $N!$ overcount. So the Einstein-solid partition function is $z^N$ with no Gibbs factor. The $1/N!$ belongs only to particles sharing one single-particle Hilbert space, like a gas.

How is this different from the Maxwell–Boltzmann classical limit article?

This article establishes the counting correction $1/N!$ — a statement about how many states there are — from the dilute limit of exact quantum combinatorics. The Maxwell–Boltzmann classical limit (aq-sm-maxwell-boltzmann-classical-limit) is about the occupation distribution: it shows the Bose–Einstein and Fermi–Dirac mean occupancies $1/(e^{\beta(\varepsilon-\mu)}\mp1)$ both reduce to $e^{-\beta(\varepsilon-\mu)}$ when the fugacity is small. They are two faces of the same classical regime $n\lambda^3\ll1$ — one about counting states, one about occupying them.

What sets the boundary of validity, numerically?

The single number $n\lambda^3$ , the phase-space density (owned by aq-sm-quantum-concentration-classicality-criterion). For air at room temperature $n\lambda^3\sim10^{-7}$ , so the $1/N!$ is spectacularly good. For conduction electrons in a metal $n\lambda^3\sim10^{3}$ , so it fails completely and Fermi–Dirac statistics are mandatory. Helium near a few kelvin sits at $n\lambda^3\sim1$ , exactly where the correction breaks down and superfluidity appears. The rule of thumb: trust $1/N!$ when $n\lambda^3\ll1$ , and reach for full quantum statistics otherwise.