Configuration Integral and the Interacting Partition Function

Two gases at the same temperature and pressure — argon and helium, say — behave measurably differently the moment you push them off the ideal-gas law. The deviation is not in how fast their molecules move; the momentum distribution is universal, fixed by temperature alone. It lives entirely in where the molecules sit relative to one another, in the interplay of hard-core repulsion and weak attraction. Classical statistical mechanics captures this with a single, decisive move: the partition function of an interacting gas splits cleanly into a kinetic part it shares with the ideal gas and a spatial part — the configuration integral $Q_N$ — that carries every consequence of the forces between molecules. The split is exact. Because the Hamiltonian adds a kinetic term (depending only on momenta) and a potential term (depending only on positions), the phase-space integral factorizes. The momentum integrals are Gaussian and collapse into powers of the thermal de Broglie wavelength $\lambda$ , exactly as for the ideal gas. What remains is $Q_N=\int e^{-\beta U}\,\mathrm{d}^{3N}r$ , and the whole partition function is $Z_N = Q_N/(N!\,\lambda^{3N})$ . Set the interactions to zero and $Q_N$ collapses to $V^N$ , returning the ideal gas; every correction to ideality is measured by how far $Q_N$ strays from $V^N$ . This article establishes that factorization and shows why $Q_N$ is the natural hub of interacting-gas theory. Rewriting the Boltzmann factor as a product of Mayer bonds turns $Q_N$ into a sum over graphs, which the Mayer cluster expansion (its own article) resums into the virial series and, in mean-field form, the van der Waals equation. Master the configuration integral and you hold the single object from which all of imperfect-gas thermodynamics radiates.

Key takeaways

The classical canonical partition function of an interacting gas factorizes exactly into a kinetic part (shared with the ideal gas) and a configuration integral $Q_N=\int e^{-\beta U}\,\mathrm{d}^{3N}r$ .
The momentum integrals are Gaussian and yield the thermal de Broglie wavelength $\lambda=h/\sqrt{2\pi m k_BT}$ ; all interaction physics lives in the single spatial integral $Q_N$ .
For a non-interacting gas $Q_N=V^N$ , recovering $Z_N=V^N/(N!\,\lambda^{3N})$ and $PV=Nk_BT$ ; deviations of $Q_N$ from $V^N$ measure non-ideality.
Pairwise additivity plus the substitution $e^{-\beta u}=1+f$ turns $Q_N$ into a sum over graphs of Mayer bonds — the seed of the cluster and virial expansions.
Because $\ln Z_N=\ln Q_N-\ln(N!\,\lambda^{3N})$ , only $\ln Q_N$ carries interactions; the configurational pressure is $P=k_BT\,\partial_V\ln Q_N$ .
Key idea: kinetic and configurational degrees of freedom decouple in classical statistical mechanics — $Z_N=(N!\,\lambda^{3N})^{-1}Q_N$ — so the entire physics of intermolecular forces is packed into one spatial integral $Q_N=\int e^{-\beta U}\,\mathrm{d}^{3N}r$ .

The core insight

Kinetic and configurational degrees of freedom decouple in classical statistical mechanics: $Z_N=(N!\,\lambda^{3N})^{-1}\,Q_N$ , so all interaction physics lives in the single spatial integral $Q_N=\int e^{-\beta U}\,\mathrm{d}^{3N}r$ .

Where this lives today

The configurational integral is the starting point of essentially every equilibrium theory of classical interacting fluids. Because $Q_N$ cannot be evaluated exactly for realistic potentials, its two great descendants dominate practice: the virial expansion computes $\ln Q_N$ order by order in density for dilute gases, and computer simulation (Monte Carlo and molecular dynamics) samples the configurational weight $e^{-\beta U}$ directly for dense liquids and solids, where the density expansion fails. Every liquid-state integral-equation theory — Percus–Yevick, hypernetted chain — is a scheme for approximating the same configurational average. In soft-matter and colloid science, effective one-component configuration integrals with density-dependent potentials describe suspensions of polymers, proteins, and nanoparticles. The deepest legacy is conceptual. The split $Z_N=Z_{\text{kinetic}}\times(Q_N/V^N)$ — universal kinetics times a configurational correction — is the classical prototype of the linked-cluster and diagrammatic expansions that pervade quantum many-body theory and field theory, where "only connected diagrams contribute to $\ln Z$ " is the exact echo of Mayer's connected-cluster theorem. Whenever a modern calculation separates a trivial reference system from a systematically expandable correction, it is walking the path the configuration integral laid down.

The physical picture

Think of the partition function as a weighted census of microstates, where each state carries the Boltzmann weight $e^{-\beta H}$ . In classical mechanics this weight factorizes because the Hamiltonian is the sum of kinetic and potential energies, $H=K+U$ , so $e^{-\beta H}=e^{-\beta K}e^{-\beta U}$ . The momentum integral is therefore the same Gaussian integral as for the ideal gas, depending only on the particle mass and temperature through the thermal de Broglie wavelength $\lambda$ . Once $T$ is fixed, all effects of intermolecular interactions are contained in how the molecules arrange themselves in space. That arrangement is exactly what $Q_N$ counts. The first figure makes the split visible: the full partition function is a kinetic box (universal, ideal-gas) multiplied by a configurational integral over all the ways $N$ particles can be positioned in the volume, weighted by $e^{-\beta U}$ . The Boltzmann factor acts as a configurational filter: arrangements with deep overlaps (hard repulsion, $U\to+\infty$ ) are exponentially suppressed toward zero weight, while arrangements sitting in attractive wells ( $U<0$ ) are mildly favored. $Q_N$ is the configurational integral obtained by summing these statistical weights over all particle positions. For a non-interacting gas, where $U=0$ , it reduces exactly to $V^N$ .

Figure

Factorization of the interacting partition function: a universal kinetic box (thermal wavelength λ) times the configuration integral over all particle positions weighted by the Boltzmann factor

Because the Mayer function $f_{ij}=e^{-\beta u(r_{ij})}-1$ is nonzero only when particles $i$ and $j$ are close enough to feel each other, the reorganized weight $\prod_{i<j}(1+f_{ij})$ is mostly ones with sparse corrections. In a dilute gas, only a small fraction of particle pairs interact simultaneously, allowing the Mayer expansion to be reorganized into a systematic virial expansion in the density. That is the deep reason $Q_N$ leads naturally to the virial series: the small parameter is $n\lambda^3$ or, more physically, the fraction of the volume actually occupied by interaction zones. For argon at ordinary conditions, both the quantum degeneracy parameter $n\lambda^3$ and the density expansion parameter are small. The first ensures that Maxwell--Boltzmann statistics is valid, while the second ensures that the virial expansion converges rapidly.

Figure

The Boltzmann factor as a configurational filter: hard-core overlaps are exponentially suppressed toward zero weight, attractive wells are mildly favored, and Q_N reduces to V^N when the potential vanishes

The thermodynamic payoff is immediate. Since $F=-k_BT\ln Z_N$ and $\ln Z_N=\ln Q_N-\ln(N!\lambda^{3N})$ , every contribution arising from intermolecular interactions enters through $\ln Q_N$ . Differentiating $\ln Q_N$ with respect to volume gives the configurational pressure that corrects the ideal-gas law. Everything downstream — virial coefficients, the van der Waals equation, the law of corresponding states — is a systematic way of computing $\ln Q_N$ order by order in density.

The core equations

The starting point is the classical canonical partition function written as a phase-space integral. The Gibbs factor $1/N!$ removes the overcounting of identical particles, while $h^{3N}$ supplies the phase-space cell that renders the partition function dimensionless. It counts every microstate weighted by its Boltzmann factor.

Z_N = \frac{1}{N!\,h^{3N}}\int \mathrm{d}^{3N}p\,\mathrm{d}^{3N}r\; e^{-\beta H(p,r)} \qquad (1)

where: $Z_N$ — canonical partition function of $N$ particles; $h$ — Planck's constant, setting the phase-space cell size; $\beta=1/(k_BT)$ ; $H=\sum_i p_i^2/2m + U(r)$ ; $N!$ — the Gibbs indistinguishability factor. Read it as a weighted sum over classical microstates: the exponential suppresses high-energy configurations, and the prefactor makes entropy extensive and the count quantum-consistent. The Hamiltonian splits into a momentum-only kinetic part and a coordinate-only potential part, so the Boltzmann factor factorizes and the momentum integrals — pure Gaussians — separate out. Doing them turns the $h^{-3N}$ into $\lambda^{-3N}$ and leaves the spatial integral untouched.

Z_N = \frac{1}{N!\,\lambda^{3N}}\,Q_N,\qquad \lambda = \frac{h}{\sqrt{2\pi m k_B T}} \qquad (2)

where: $Q_N$ — the configuration integral (below); $\lambda$ — thermal de Broglie wavelength, the quantum length scale of a thermal particle; $m$ — particle mass. The physics: temperature and mass fix the kinetic contribution completely and identically for every gas, so the entire chemical identity of the gas — how it deviates from ideal — is displaced into $Q_N$ . The configuration integral itself is the integral of the Boltzmann factor of the interaction energy over all particle positions. It is the one object in the theory that resists exact evaluation for a real potential.

Q_N = \int_V \mathrm{d}^{3N}r\; e^{-\beta U(\mathbf{r}_1,\dots,\mathbf{r}_N)} \qquad (3)

where: $U$ — total potential energy of the configuration; $\int_V$ — each of the $N$ coordinates ranges over the volume. When $U=0$ the integrand is $1$ and $Q_N=V^N$ ; every departure from $V^N$ is a fingerprint of the forces. Almost all realistic intermolecular energy is pairwise additive: the total is a sum over distinct pairs of a two-body potential depending only on separation. This is the assumption that makes the whole graphical machinery possible.

U(\mathbf{r}_1,\dots,\mathbf{r}_N) = \sum_{i<j} u(r_{ij}),\qquad r_{ij}=|\mathbf{r}_i-\mathbf{r}_j| \qquad (4)

where: $u(r)$ — the pair potential (Lennard-Jones, hard sphere, ...); $r_{ij}$ — separation of particles $i$ and $j$ . Three-body and higher forces exist but are typically small for closed-shell atoms; pairwise additivity is the workhorse approximation of imperfect-gas theory. With pairwise additivity the exponential of a sum becomes a product over pairs, and Mayer's substitution $e^{-\beta u}=1+f$ replaces each long-ranged Boltzmann factor by a short-ranged bond. This is the pivot to the cluster expansion, whose own article (aq-sm-mayer-f-function) owns the $f$ -function in detail.

e^{-\beta U} = \prod_{i<j}\big(1+f_{ij}\big),\qquad f_{ij}=e^{-\beta u(r_{ij})}-1 \qquad (5)

where: $f_{ij}$ — the Mayer $f$ -function, which vanishes at large separation. Here the product is mostly ones, so expanding it generates a controlled, sparse sum — the graph expansion that aq-sm-mayer-cluster-expansion resums. Finally, the thermodynamics. The free energy separates into the ideal-gas contribution plus an interaction correction through $\ln(Q_N/V^N)$ . Consequently, the pressure is the ideal-gas pressure plus a configurational correction obtained from the volume dependence of $Q_N$ .

F = F_{\text{ideal}} - k_B T\,\ln\!\frac{Q_N}{V^N},\qquad P = k_B T\left(\frac{\partial \ln Q_N}{\partial V}\right)_{T,N} \qquad (6)

where: $F_{\text{ideal}}=-k_BT\ln\!\big(V^N/N!\,\lambda^{3N}\big)$ ; $P$ — pressure. Because only $\ln Q_N$ depends on the interactions, $\ln(Q_N/V^N)$ is exactly the excess free energy over ideal — the number the virial expansion computes.

Derivation AFactorizing the partition function into kinetic and configurational parts

The goal is to prove the exact split $Z_N=Q_N/(N!\,\lambda^{3N})$ starting from the phase-space integral (1), and in doing so to show precisely where the thermal de Broglie wavelength comes from and why the configuration integral inherits all of the interaction physics. The route is elementary but worth doing carefully once: separate the Boltzmann factor because the Hamiltonian is a sum, perform the $3N$ Gaussian momentum integrals, and convert the leftover $h^{-3N}$ into $\lambda^{-3N}$ . Watch how the potential $U(r)$ never touches the momentum integrals, so it survives untouched inside $Q_N$ — that is the mathematical statement that kinetics and configuration decouple.

Step 1. Split the Boltzmann factor. Since $H=\sum_i p_i^2/2m + U(\mathbf r)$ , the exponential separates: $e^{-\beta H}=e^{-\beta\sum_i p_i^2/2m}\,e^{-\beta U(\mathbf r)}$ . Why: the exponent is a sum of a momentum-only term and a coordinate-only term, and the exponential of a sum is a product — the potential factor simply rides along, independent of $p$ .

Step 2. Do one Gaussian momentum integral. For a single Cartesian component, $\int_{-\infty}^{\infty}\mathrm{d}p\,e^{-\beta p^2/2m}=\sqrt{2\pi m/\beta}=\sqrt{2\pi m k_BT}$ . Why: this is the standard Gaussian $\int e^{-a p^2}\mathrm{d}p=\sqrt{\pi/a}$ with $a=\beta/2m$ ; the three components per particle give the cube, and there are $N$ particles.

Step 3. Collect all $3N$ momentum integrals and form $\lambda$ . The full momentum factor is $(2\pi m k_BT)^{3N/2}$ ; dividing by $h^{3N}$ gives $\big(\sqrt{2\pi m k_BT}/h\big)^{3N}=\lambda^{-3N}$ with $\lambda=h/\sqrt{2\pi m k_BT}$ . Why: each factor of $h$ pairs with a $\sqrt{2\pi m k_BT}$ to build one inverse thermal wavelength; $\lambda$ is literally the momentum integral measured in units of $h$ .

Step 4. Read off the spatial remainder. What survives is $Z_N=\dfrac{1}{N!\,\lambda^{3N}}\displaystyle\int_V \mathrm{d}^{3N}r\,e^{-\beta U}=\dfrac{Q_N}{N!\,\lambda^{3N}}$ . Why: the position integrals were never touched by the momentum manipulation, so they assemble into $Q_N$ exactly; the $N!$ from indistinguishability is carried along unchanged. *As a consistency check: the identity $(1/h^3)(2\pi m k_BT)^{3/2}=\lambda^{-3}$ follows directly from the Gaussian integrals above, and evaluating it for argon at $300\,\mathrm{K}$ gives $\lambda(\mathrm{Ar},300\,\mathrm{K})=1.60\times10^{-11}$ m, consistent with the worked example in the next section.*

Go deeper · Derivation B · 1 minThe ideal-gas limit and the configurational pressure

Derivation BThe ideal-gas limit and the configurational pressure

Here the goal is to show two things at once: that setting $U=0$ collapses $Q_N$ to $V^N$ and reproduces the ideal-gas law $PV=Nk_BT$ , and that when $U\neq0$ the pressure is governed entirely by $\ln Q_N$ through $P=k_BT\,\partial_V\ln Q_N$ . The strategy is to compute $Q_N$ in the ideal limit directly, take the logarithm of $Z_N$ , and differentiate with respect to volume — carefully tracking which terms carry the $V$ dependence. The point to watch is that $\lambda$ has no volume dependence, so in the ideal case the entire pressure comes from the explicit $V^N$ ; once interactions turn on, the extra $V$ -dependence hides inside $Q_N$ , which is why the configurational pressure is a derivative of $\ln Q_N$ .

Step 1. Evaluate $Q_N$ for the ideal gas. With $U=0$ the integrand is $1$ , so $Q_N=\int_V\mathrm{d}^{3N}r=V^N$ — each of the $N$ position integrals independently gives the volume. Why: nothing weights one configuration over another when there are no forces, so the "configurational volume" is just the geometric volume raised to the number of particles.

Step 2. Form the ideal free energy. Then $Z_N=V^N/(N!\,\lambda^{3N})$ and $F_{\text{ideal}}=-k_BT\ln Z_N=-k_BT\big[N\ln V-\ln N!-3N\ln\lambda\big]$ . Why: this is exactly the Sackur–Tetrode-generating free energy; the $N!$ is what makes $F$ extensive (the Gibbs resolution of the mixing paradox).

Step 3. Differentiate for the pressure. $P=-(\partial F/\partial V)_{T,N}=k_BT\,\partial_V(N\ln V)=Nk_BT/V$ , i.e. $PV=Nk_BT$ . Why: only the $N\ln V$ term depends on $V$ ; $\lambda$ and $N!$ are volume-independent, so the ideal-gas law drops out with no interaction physics at all.

Step 4. Restore interactions. For $U\neq0$ , $\ln Z_N=\ln Q_N-\ln(N!\,\lambda^{3N})$ , or equivalently $\ln Z_N = N\ln V - \ln(N!\,\lambda^{3N}) + \ln\!\left(\frac{Q_N}{V^N}\right)$ . Differentiating gives $P = \dfrac{Nk_BT}{V} + k_BT\left(\dfrac{\partial}{\partial V}\ln\dfrac{Q_N}{V^N}\right)_{T,N}$ . Why: the first term is the ideal-gas pressure, while the second is the excess (configurational) pressure arising from intermolecular interactions. *As a self-consistency check: $Q_N=V^N$ for $U=0$ , the ideal-gas pressure $P=Nk_BT/V$ from $Z_N=V^N/(N!\,\lambda^{3N})$ , and the free-energy split $\ln Z_N=\ln Z_{\text{ideal}}+\ln(Q_N/V^N)$ all follow directly from the steps above.*

Limiting cases

The configuration integral interpolates between the ideal gas and the fully interacting fluid; the controlling parameter is the density (through the fraction of configurations with overlapping interaction zones).

Case	Result	Meaning
No interaction ( $U=0$ )	$Q_N=V^N$ , $Z_N=V^N/(N!\,\lambda^{3N})$	Recovers the ideal gas and $PV=Nk_BT$ ; the configurational filter is transparent.
Dilute gas ( $n\lambda^3\ll1$ , few pairs close)	$Q_N/V^N = 1 + O(n)$	The product $\prod(1+f_{ij})$ is mostly $1$ ; corrections organize into the virial series in powers of density.
Hard spheres	$Q_N/V^N<1$	Excluded volume removes overlapping configurations, lowering $Q_N$ below $V^N$ ; pressure rises above ideal.
Attraction-dominated (low $T$ )	$Q_N/V^N>1$ (locally)	Attractive wells favor near configurations, raising $Q_N$ and lowering the pressure below ideal.
Dense liquid / near condensation	$Q_N$ non-expandable	Many pairs overlap at once; the density expansion breaks down and $Q_N$ must be treated non-perturbatively (simulation, integral equations).

Common trapThe 1/N! is a quantum indistinguishability effect, so a purely classical configuration integral does not need it

The $1/N!$ is not optional and it is not "quantum" in the dynamical sense. It is the Gibbs correction that makes the entropy extensive and resolves the Gibbs paradox — a purely combinatorial statement that the $N$ particles carry no labels, so permuting them does not produce a new microstate. Drop it and the free energy fails to be additive: joining two identical boxes would spuriously change the entropy. The configuration integral $Q_N=\int e^{-\beta U}\,\mathrm{d}^{3N}r$ deliberately integrates over labeled positions and overcounts by exactly $N!$ ; the prefactor divides that overcount out. A second, subtler trap: $Q_N$ has units of $V^N$ (volume $^N$ ), so it is not dimensionless — only the ratio $Q_N/V^N$ and the full $Z_N$ (with $\lambda^{3N}$ ) are the physically meaningful, dimensionless quantities.

Worked examples

1 · The thermal wavelength and the diluteness of argon (confidence builder)

We compute the thermal de Broglie wavelength of argon at room temperature and the diluteness parameter $n\lambda^3$ , to see quantitatively why the configuration-integral expansion is so well controlled for a real gas.

Step 1. Argon has mass $m=39.95\,u=6.63\times10^{-26}\,\mathrm{kg}$ . At $T=300$ K, $\lambda=\dfrac{h}{\sqrt{2\pi m k_BT}}=\dfrac{6.626\times10^{-34}}{\sqrt{2\pi(6.63\times10^{-26})(1.381\times10^{-23})(300)}}$ . The denominator is $\sqrt{2\pi\times6.63\times10^{-26}\times4.14\times10^{-21}}=\sqrt{1.726\times10^{-45}}=4.15\times10^{-23}$ , so $\lambda=\mathbf{1.60\times10^{-11}\ m}$ — about $0.16$ Å, far smaller than the $\sim3.4$ Å molecular size.

Step 2. At standard conditions the number density is $n=2.5\times10^{25}\,\mathrm{m^{-3}}$ . The diluteness parameter is $n\lambda^3 = 2.5\times10^{25}\times(1.60\times10^{-11})^3 = 2.5\times10^{25}\times4.06\times10^{-33} = \mathbf{1.0\times10^{-7}}$ .

Step 3. Because $n\lambda^3\approx10^{-7}\ll1$ , quantum-statistical effects are completely negligible and the classical configurational integral is valid. Independently, ordinary argon is also a dilute fluid, so the virial expansion of $Q_N/V^N$ converges rapidly because the density (or packing fraction) is small. These are distinct approximations: $n\lambda^3\ll1$ justifies Maxwell--Boltzmann statistics, whereas low density controls the configurational expansion.

2 · How the configuration integral shifts the pressure (exam level)

We show, to leading order in density, how $Q_N$ produces a pressure correction, using a schematic single-pair estimate to connect $\ln Q_N$ to the second virial coefficient owned by later articles.

Step 1. Expand the product to first order in Mayer bonds: $Q_N=\int\prod_{i<j}(1+f_{ij})\,\mathrm{d}^{3N}r \approx V^N + V^{N-2}\!\!\sum_{i<j}\int f_{ij}\,\mathrm{d}^3r_i\,\mathrm{d}^3r_j$ . There are $\binom{N}{2}\approx N^2/2$ pairs, and each bond integral is $V\int f(r)\,\mathrm{d}^3r$ after using translation invariance.

Step 2. So $Q_N/V^N \approx 1 + \dfrac{N^2}{2V}\int f(r)\,\mathrm{d}^3r$ , giving $\ln(Q_N/V^N)\approx \dfrac{N^2}{2V}\int f(r)\,\mathrm{d}^3r$ to leading order.

Step 3. The pressure correction is $P-P_{\text{ideal}} = k_BT\,\partial_V\ln(Q_N/V^N) = -k_BT\dfrac{N^2}{2V^2}\int f(r)\,\mathrm{d}^3r$ . Writing $n=N/V$ and identifying $B_2=-\tfrac12\int f(r)\,\mathrm{d}^3r$ (owned by aq-sm-second-virial-coefficient), this is $\dfrac{P}{nk_BT} = 1 + B_2\,n + \cdots$ .

Step 4. For argon-like Lennard-Jones parameters ( $\varepsilon/k_B=120$ K, $\sigma=0.34$ nm) at $300$ K, the numerics give $\int f\,\mathrm{d}^3r = 5.5\times10^{-29}\,\mathrm{m^3}$ and hence $\mathbf{B_2\approx-16\ cm^3/mol}$ — negative, meaning attraction wins at this temperature and the real gas is slightly more compressible than ideal. The configuration integral has converted the microscopic intermolecular potential into a measurable correction to the equation of state. One systematic way of evaluating it is the Mayer cluster expansion, developed in the following articles.

Origins · 1 minJoseph Edward Mayer and how this was discovered

Joseph Edward Mayer

1904–1983 · American physical chemist who built the cluster expansion of the imperfect gas

Joseph Edward Mayer transformed the intractable configuration integral into a systematic, computable theory of the imperfect gas. Working in the 1930s, he introduced the substitution $e^{-\beta u}=1+f$ — trading the long-ranged Boltzmann factor for the short-ranged Mayer function $f$ — and showed that the configuration integral could be reorganized as a sum over cluster diagrams, with the logarithm of the grand partition function keeping only the connected ones. This linked-cluster structure turned intermolecular potentials into virial coefficients that could be compared directly with experiment, and it became the classical ancestor of every diagrammatic linked-cluster method in later many-body physics and quantum field theory. With his wife Maria Goeppert Mayer he wrote the classic 1940 textbook Statistical Mechanics, which codified the whole program. (Maria Goeppert Mayer went on to win the 1963 Nobel Prize in Physics — not for the cluster expansion, but for the nuclear shell model.) The configuration integral defined here is the object Mayer's expansion tames.

History

1901

— Gibbs's Elementary Principles in Statistical Mechanics frames the partition function and the ensemble formalism.

1902

— Gibbs introduces the indistinguishability

1/N!

that makes the entropy extensive.

1927

— Ursell introduces cluster (Ursell) functions for the classical gas, anticipating the graph expansion.

1937

— Joseph Mayer develops the cluster expansion and the Mayer

f

-function for imperfect gases.

1940

— Joseph and Maria Goeppert Mayer publish Statistical Mechanics, codifying the configuration-integral machinery.

Going further

Texts. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 9–10 (the interacting classical gas, configuration integral, and cluster expansion). Kardar, Statistical Physics of Particles, Ch. 5. Hansen & McDonald, Theory of Simple Liquids (the liquid-state descendants). McQuarrie, Statistical Mechanics, Ch. 12. The key numeric and symbolic results — the thermal wavelength, the $Q_N=V^N$ ideal limit, the ideal-gas pressure, and the free-energy split — are all consistent with one another and with the derivations presented in this article.

Prerequisites. Canonical partition function · Classical phase space and Gibbs factor · The Mayer f-function.

Leads to. Mayer cluster expansion · Cluster integrals · Virial expansion · Second virial coefficient.

Why do we divide by N! and by λ^(3N) but not by anything else?

The $N!$ removes the overcounting from treating identical particles as labeled — it is the Gibbs correction that makes entropy extensive. The $\lambda^{3N}$ (equivalently the $h^{3N}$ in the phase-space integral) makes the partition function dimensionless and quantum-consistent: each particle occupies a phase-space cell of size $h^3$ , and $\lambda^3$ is that cell measured in configuration space at temperature $T$ . There is nothing else to divide by because the split is exact: once the Gaussian momentum integrals are done, the only surviving object is the spatial integral $Q_N$ , and $N!\,\lambda^{3N}$ is precisely the ideal-gas normalization it sits inside.

Is Q_N a probability or a volume? Its units confuse me.

$Q_N$ has units of volume $^N$ (each of the $N$ position integrals contributes a factor of volume). It is best read as a weighted configurational volume: for an ideal gas it is exactly $V^N$ , the total geometric volume available to $N$ independent particles, and interactions reweight it up or down. It is not a probability. The dimensionless, physically meaningful quantities are the ratio $Q_N/V^N$ (the excess over ideal) and the full partition function $Z_N=Q_N/(N!\,\lambda^{3N})$ , which is a pure number.

What exactly breaks down when the gas gets dense?

The expansion of $\prod_{i<j}(1+f_{ij})$ assumes that only a few pairs are within interaction range at once, so that most bonds are "off" ( $f=0$ ) and the corrections are sparse. At high density many particles overlap simultaneously, so a large fraction of bonds are "on," and the density series (the virial expansion of $\ln Q_N$ ) converges slowly or not at all — especially near condensation and at the critical point. There, $Q_N$ must be attacked non-perturbatively: molecular-dynamics or Monte-Carlo sampling of $e^{-\beta U}$ , or integral-equation closures for the pair correlation function. The configuration integral is still the right object; only the expansion of it fails.

Does the factorization survive if there are three-body forces?

The kinetic–configurational split (equation 2) survives regardless of the form of $U$ , because it relies only on $H=K(p)+U(r)$ — the potential can be anything, three-body or worse, and the momentum integrals still separate to give $\lambda^{3N}$ . What changes is the pairwise-additivity step (equation 4) and Mayer's $e^{-\beta U}=\prod(1+f_{ij})$ : those assume $U=\sum_{i<j}u(r_{ij})$ . With genuine three-body forces you must add three-body Mayer functions and extra diagram types. For closed-shell atoms three-body terms (Axilrod–Teller) are usually a few percent, so pairwise additivity is an excellent, and standard, starting point.

Why is the pressure a derivative of ln Q_N and not of Q_N itself?

Because pressure comes from the free energy, $P=-(\partial F/\partial V)_{T,N}$ , and $F=-k_BT\ln Z_N$ . The logarithm is what makes thermodynamic quantities intensive and additive: $\ln Z_N=\ln Q_N-\ln(N!\,\lambda^{3N})$ , and only $\ln Q_N$ carries the $V$ -dependence from interactions. Differentiating $\ln Q_N$ rather than $Q_N$ automatically produces the intensive configurational pressure and correctly reduces to $Nk_BT/V$ in the ideal limit. Working with $Q_N$ directly would tangle the exponentially large ideal-gas factor into every step.

How does this connect to the van der Waals equation I learned in chemistry?

The van der Waals equation is what you get by evaluating $\ln Q_N$ to first order in density (the second virial coefficient) with a simple potential — a hard core plus a weak attractive tail — and then resumming that leading correction in a mean-field way. The hard core gives the excluded-volume constant $b$ (the positive, repulsive part of $B_2$ ), and the attraction gives the constant $a$ (the negative part). So van der Waals is not a separate theory: it is the configuration integral, truncated at the second virial coefficient and dressed up. The route runs through aq-sm-mayer-cluster-expansion and aq-sm-second-virial-coefficient.

Does the Gibbs 1/N! ever cause trouble for distinguishable particles, like a mixture?

Yes — for a mixture you use a product of factorials, one per species: $1/(N_A!\,N_B!\cdots)$ , because particles of different species are distinguishable and must not be de-counted against each other. This is exactly what makes the entropy of mixing of distinct gases nonzero while the "mixing" of identical gases gives zero (the Gibbs paradox resolution). For a one-component gas the single $1/N!$ is correct; for interacting mixtures the configuration integral itself also acquires cross-species Mayer bonds $f_{AB}$ .