The Mayer f-Function

Try to build a systematic theory of a real gas by expanding in the interaction potential and you hit an immediate wall. The configurational Boltzmann weight is a product of pair factors $e^{-\beta u(r_{ij})}$ , and each of those factors tends to one — not zero — when the particles are far apart. A factor that stays order-one everywhere is useless as a small parameter: the naive power series in $u$ diverges term by term for a hard core, and the product refuses to break into independent pieces. The gas seems to couple all $N$ particles at once, and the configuration integral looks hopeless. Mayer's decisive trick, introduced in 1937, is to subtract the trivial value. Write $e^{-\beta u_{ij}} = 1 + f_{ij}$ with the Mayer $f$ -function $f_{ij}=e^{-\beta u(r_{ij})}-1$ . The new object vanishes wherever the particles are far apart — precisely where the old factor was uselessly close to one. It is bounded between $-1$ (deep in a hard core, where the Boltzmann factor is zero) and a modest positive value (in an attractive well), so it is integrable over all space even for a singular potential. And its sign carries physics on its sleeve: $f<0$ where the potential repels, $f>0$ where it attracts. Because $f_{ij}$ has finite range, the product $\prod_{i<j}(1+f_{ij})$ is mostly ones, and expanding it generates a sum of terms — each a product of $f$ -bonds — that maps one-to-one onto graphs drawn on the particle labels. That graphical reorganization is the entire start of the Mayer cluster expansion (its own article): connected clusters survive in $\ln\Xi$ , the single-bond cluster gives the second virial coefficient, and higher clusters give the full equation of state. The Mayer $f$ -function is thus the pivot between a microscopic potential and measurable thermodynamics, and this article owns it: its definition, its limits, its sign structure, and the graph expansion it seeds.

Key takeaways

The Mayer $f$ -function $f_{ij}=e^{-\beta u(r_{ij})}-1$ replaces the bare pair Boltzmann factor by a quantity that decays to zero once the separation exceeds the interaction range.
Because $f_{ij}\to0$ as $r_{ij}\to\infty$ , every distant particle pair contributes the factor $1$ , and expanding $\prod_{i<j}(1+f_{ij})$ therefore generates a sum over graphs in which only interacting pairs are connected by Mayer bonds.
For a typical potential $f\to-1$ in the hard core and rises to a small positive bump in the attractive well, so $f$ is bounded and integrable even where $u$ diverges.
The integral $-\tfrac12\int f\,\mathrm{d}^3r$ is exactly the second virial coefficient $B_2$ , so $f$ is the microscopic input to the virial expansion (owned by later articles).
The $f$ -function is the building block of the Mayer cluster expansion and the whole low-density theory of imperfect gases.
Key idea: subtract the trivial value — $e^{-\beta u}=1+f$ with $f=e^{-\beta u}-1$ vanishing at large $r$ — so products of pair Boltzmann factors expand into a controlled, integrable sum over graphs.

The core insight

Subtract the trivial value: $e^{-\beta u}=1+f$ with $f=e^{-\beta u}-1$ vanishing at large $r$ , so products of Boltzmann factors expand into a controlled, integrable graph sum.

Where this lives today

The Mayer $f$ -function is the fundamental microscopic ingredient in equilibrium theories of gas non-ideality and underlies the calculation of virial coefficients and related thermodynamic corrections. Tabulated second virial coefficients — used to correct real-gas thermometry, to calibrate reference equations of state, and to infer interatomic potentials from $B_2(T)$ data — are all $-\tfrac12\int f\,\mathrm{d}^3r$ evaluated for the appropriate potential. Modern ab initio pair potentials for helium and other light gases are so accurate that the computed $B_2(T)$ now serves as a primary pressure and temperature standard in metrology, closing the loop from quantum chemistry to the definition of the kelvin. Beyond dilute gases, the same bonded-graph structure recurs everywhere a reference system is corrected by a perturbation. In liquid-state theory the $f$ -bond generalizes to the direct correlation function and the diagrammatic expansions behind the Percus–Yevick and hypernetted-chain closures. In quantum many-body physics and quantum field theory, the same linked-cluster principle reappears as the statement that only connected diagrams contribute to $\ln Z$ , providing the organizing principle of perturbative expansions. The simple decomposition $f=e^{-\beta u}-1$ is the classical prototype of the connected-diagram expansions that organize much of modern statistical mechanics and quantum many-body theory.

The physical picture

Picture the gas as a set of points. The exact configurational weight is a product over all pairs, $\prod_{i<j}e^{-\beta u_{ij}}$ . Expanding the Boltzmann factors directly is unhelpful because every pair contributes a factor of order unity, even when the particles are far apart and effectively non-interacting — two particles a micron apart still contribute a factor of $1$ , and there are astronomically many such "nothing" factors. The $f$ -function strips away the "nothing happening" part: $f_{ij}$ is literally zero unless particles $i$ and $j$ are close enough to feel each other. The Mayer function removes the trivial long-range contribution, so only genuinely interacting particle pairs contribute to the expansion. The first figure shows the whole story of a single pair. Take the Lennard-Jones potential $u(r)=4\varepsilon[(\sigma/r)^{12}-(\sigma/r)^6]$ — a steep repulsive wall and a shallow attractive well. Below $r=\sigma$ the potential rises fast and $f$ plunges to $-1$ ; at $r=\sigma$ the potential crosses zero and $f=0$ ; at the minimum $r_{\min}=2^{1/6}\sigma$ the potential is $-\varepsilon$ and $f=e^{\beta\varepsilon}-1>0$ ; and at large $r$ both $u$ and $f$ decay to zero. The punchline is the contrast: $u$ diverges at small $r$ , but $f$ merely saturates at $-1$ — the $f$ -function is bounded and integrable where the potential is not.

Figure

The Lennard-Jones potential u(r) diverges at small r but the Mayer function f(r) saturates at −1: u crosses zero at r=σ (where f=0), dips to −ε at r_min (where f is a positive bump), and both decay to zero at large r

The sign of the Mayer function has a simple physical interpretation. Where the potential is repulsive ( $u>0$ ) we have $f<0$ : those configurations are suppressed relative to no interaction, because it costs energy to overlap. Where the potential is attractive ( $u<0$ ) we have $f>0$ : those configurations are favored, because sitting in the well lowers the energy. The integral of $f$ then measures the net balance of excluded versus attracted volume, which is precisely what shifts the pressure away from the ideal-gas value — and, up to a factor of $-\tfrac12$ , it is the second virial coefficient. The second figure shows why the finite range converts products into graphs. Because $f_{ij}$ is nonzero only for close pairs, a product like $f_{12}f_{34}$ requires two independent close pairs, while $f_{12}f_{23}$ requires a connected chain $1$ – $2$ – $3$ . Each product of bonds is a graph; short range means distant subgraphs decouple. This is exactly why the configuration integral reduces to counting connected clusters — the bookkeeping the cluster expansion completes.

Figure

Products of Mayer bonds are graphs: f₁₂f₃₄ is two independent close pairs (disconnected), while f₁₂f₂₃ is a connected chain 1–2–3; short range means distant subgraphs decouple

The core equations

The definition is a single subtraction, but it is the one that makes everything else work. The bare pair Boltzmann factor is replaced by its deviation from unity.

f_{ij} \equiv e^{-\beta u(r_{ij})} - 1 \qquad (1)

where: $f_{ij}$ — the Mayer function for the pair $(i,j)$ ; $u(r_{ij})$ — the pair potential at separation $r_{ij}=|\mathbf r_i-\mathbf r_j|$ ; $\beta=1/(k_BT)$ . Physically, $f_{ij}$ measures the change in the pair Boltzmann weight produced by the interaction, relative to two non-interacting particles: zero when they cannot feel each other, negative when they repel, positive when they attract. The whole point is the rearranged pair factor: each Boltzmann factor is one plus a short-ranged correction, so the total weight becomes a product over $(1+f)$ bonds.

e^{-\beta u_{ij}} = 1 + f_{ij},\qquad e^{-\beta\sum_{i<j} u_{ij}} = \prod_{i<j}\left(1+f_{ij}\right) \qquad (2)

where: the left identity is just the definition rearranged; the right follows from pairwise additivity $U=\sum_{i<j}u_{ij}$ . Because $f_{ij}\to0$ as $r_{ij}\to\infty$ , distant particle pairs contribute the trivial factor $1$ . Expanding $\prod_{i<j}(1+f_{ij})$ therefore generates a sum over products of Mayer bonds, each corresponding to a graph on the particle labels. Expanding the product assigns either the factor $1$ or the Mayer bond $f_{ij}$ to every particle pair, thereby generating every graph on the $N$ labelled vertices — one graph for each choice of which pairs contribute an $f$ -bond rather than a $1$ .

\prod_{i<j}(1+f_{ij}) = 1 + \sum_{i<j} f_{ij} + \sum_{\substack{i<j,\,k<l\\ \text{distinct pairs}}} f_{ij}f_{kl} + \cdots = \sum_{G} \prod_{(i,j)\in G} f_{ij} \qquad (3)

where: $G$ runs over all graphs on the $N$ vertices; $\prod_{(i,j)\in G}f_{ij}$ is the product of bonds on the present edges. The empty graph gives the ideal $1$ ; single bonds give the leading density correction; the full sum, once resummed in the grand ensemble, is the cluster expansion (aq-sm-mayer-cluster-expansion). The lowest nontrivial graph — a single bond between two particles — integrated over space gives the second virial coefficient. This is the quantity one measures, and the $f$ -function is its microscopic definition.

B_2(T) = -\frac{1}{2}\int f(r)\,\mathrm{d}^3r = -2\pi\int_0^\infty \left(e^{-\beta u(r)}-1\right) r^2\,\mathrm{d}r \qquad (4)

where: $B_2(T)$ — the second virial coefficient in $P/(nk_BT)=1+B_2 n+\cdots$ ; the $-\tfrac12$ is the pair symmetry factor. Because $f$ changes sign, $B_2$ can be positive (repulsion/excluded-volume dominant, high $T$ ) or negative (attraction dominant, low $T$ ), and it passes through zero at the Boyle temperature. The full derivation and the higher $B_n$ are owned by aq-sm-second-virial-coefficient and aq-sm-mayer-cluster-expansion; here $f$ supplies the integrand.

Derivation AFrom pairwise additivity to a sum over graphs

The goal is to show, step by step, how the exact configurational weight $e^{-\beta U}$ of a pairwise-additive gas becomes a sum over graphs of Mayer bonds, and to make clear why the finite range of $f$ is what makes this expansion useful rather than a formal rearrangement. The strategy is three clean algebraic moves — factorize the exponential of a sum, insert $e^{-\beta u}=1+f$ pair by pair, then expand the product — followed by the crucial physical reading of the result as a graph sum. The thing to watch is that the expansion is not an approximation at all: equation (3) is exact for any potential. The approximation, and the usefulness, come later, when short range lets us truncate the graph sum at low density.

Step 1. Factorize the exponential of the pairwise sum. Since $U=\sum_{i<j}u_{ij}$ is additive over pairs, $e^{-\beta U}=e^{-\beta\sum_{i<j}u_{ij}}=\prod_{i<j}e^{-\beta u_{ij}}$ . Why: the exponential of a sum is a product of exponentials — this is where pairwise additivity does its work; without it there would be no product to expand.

Step 2. Insert the Mayer substitution on every pair. Replace each factor using $e^{-\beta u_{ij}}=1+f_{ij}$ , giving $\prod_{i<j}(1+f_{ij})$ . Why: each $f_{ij}$ is bounded and short-ranged, unlike the order-one $e^{-\beta u_{ij}}$ ; we have traded a product of "always-on" factors for a product of "mostly-off" ones.

Step 3. Expand the product into monomials. With $\binom{N}{2}$ factors, choosing either $1$ or $f_{ij}$ in each gives $\prod_{i<j}(1+f_{ij})=\sum_{G}\prod_{(i,j)\in G}f_{ij}$ , where $G$ ranges over all subsets of pairs — i.e. all graphs on $N$ vertices. Why: multiplying out a product of binomials selects, for each pair, whether its bond is present; a set of present bonds is exactly a graph, so the terms are in one-to-one correspondence with graphs.

Step 4. Read the graph sum physically. A bond $f_{ij}$ is nonzero only when $i,j$ are within interaction range, so each surviving graph glues its vertices into spatially compact clusters; the empty graph is the ideal gas, a single bond is the leading correction, and connected multi-bond graphs give successive density corrections. Why: short range means a term with $k$ bonds needs $k$ simultaneous close encounters, whose probability scales with density — this is why the graph sum is secretly a density expansion. *The sign and limits of $f$ — $f\to0$ at $u\to0$ , $f\to-1$ at $u\to+\infty$ , $f>0$ for $u<0$ — together with the leading linearization $f\approx-\beta u$ and the Lennard-Jones landmark values $f(\sigma)=0$ and $f(r_{\min})=e^{\beta\varepsilon}-1$ all follow directly from the definition $f=e^{-\beta u}-1$ .*

Go deeper · Derivation B · 2 minThe second virial coefficient from a single Mayer bond

Derivation BThe second virial coefficient from a single Mayer bond

Here the goal is to extract the second virial coefficient $B_2(T)=-\tfrac12\int f\,\mathrm{d}^3r$ from the single-bond term of the graph expansion, and to show why its sign flips between low and high temperature — the microscopic origin of the Boyle temperature. The route is to keep only the one-bond graphs in the configuration integral, use translation invariance to reduce the six-dimensional pair integral to a one-dimensional radial integral, and read off the density correction to the pressure. What to watch is that this article owns the integrand $f$ but only sketches the bookkeeping to $B_2$ ; the full derivation, the symmetry factor, and the higher coefficients are handed to aq-sm-mayer-cluster-expansion and aq-sm-second-virial-coefficient by name.

Step 1. Keep the single-bond graphs. To leading order beyond ideal, $Q_N/V^N \approx 1 + \dfrac{1}{V^2}\displaystyle\sum_{i<j}\int f_{ij}\,\mathrm{d}^3r_i\,\mathrm{d}^3r_j$ , with $\binom{N}{2}\approx N^2/2$ identical pair terms. Why: two-bond and higher graphs carry extra powers of density, so at leading order only the single-bond graph survives.

Step 2. Reduce with translation invariance. Because $f_{ij}=f(|\mathbf r_i-\mathbf r_j|)$ depends only on the separation, $\int f_{ij}\,\mathrm{d}^3r_i\,\mathrm{d}^3r_j = V\int f(r)\,\mathrm{d}^3r$ ; the center-of-mass integral gives one factor of $V$ . Why: shifting both particles together does not change $f$ , so one of the two three-dimensional integrals is a free volume factor — this is what makes the cluster integral intensive.

Step 3. Collect the density correction. Then $\dfrac{Q_N}{V^N}\approx 1+\dfrac{N^2}{2V}\int f(r)\,\mathrm{d}^3r$ , and differentiating $\ln(Q_N/V^N)$ for the pressure gives $\dfrac{P}{nk_BT}=1+B_2 n+\cdots$ with $B_2=-\tfrac12\int f(r)\,\mathrm{d}^3r$ . Why: the $-\tfrac12$ is the pair symmetry factor from the $N^2/2$ pairs and the volume derivative; this is the point at which the microscopic $f$ becomes a measurable equation-of-state coefficient.

Step 4. Read the sign. With $f<0$ where $u>0$ (repulsion) and $f>0$ where $u<0$ (attraction), $B_2$ is positive when the hard core dominates (high $T$ ) and negative when attraction dominates (low $T$ ), vanishing at the Boyle temperature. Why: raising $T$ shrinks $\beta\varepsilon$ , weakening the positive attractive bump in $f$ while the hard-core $f=-1$ region is temperature-independent — so excluded volume eventually wins and $B_2$ turns positive. *Evaluating $B_2=-2\pi\int(e^{-\beta u}-1)r^2\mathrm{d}r$ for the Lennard-Jones potential gives $+0.83\,\sigma^3$ at $1000$ K (repulsion-dominated), $-7.3\,\sigma^3$ at $100$ K (attraction-dominated), and $B_2\approx-16$ cm³/mol for argon at $300$ K, in agreement with the experimental value.*

Limiting cases

The Mayer function is entirely controlled by the sign and depth of the potential relative to $k_BT$ .

Case	Result	Meaning
Large separation ( $u\to0$ )	$f\to0$	No interaction, no bond; this is why $f$ is short-ranged and integrable.
Hard core ( $u\to+\infty$ )	$f\to-1$	Boltzmann factor is zero; $f$ saturates at $-1$ rather than diverging — bounded and integrable.
Attractive well ( $u<0$ )	$f=e^{\beta\lvert u\rvert}-1>0$	Configurations favored; $f$ is a positive bump whose height grows as $T$ falls.
Weak potential ( $\beta\lvert u\rvert\ll1$ )	$f\approx-\beta u$	Linearizes in the potential; recovers the high-temperature/mean-field regime.
High $T$	$B_2>0$	Repulsion/excluded-volume dominates; gas is stiffer than ideal.
Low $T$	$B_2<0$	Attraction dominates; gas is more compressible than ideal, heading toward condensation.

Common trapThe Mayer function is small, so it is just a linearized potential −βu

Only when the potential is weak. In general $f=e^{-\beta u}-1$ is a fully nonlinear function of $u$ , and it is not small everywhere — it equals $-1$ throughout the hard core, where $\beta u$ is enormous, and it is the nonlinearity that keeps $f$ bounded there. Replacing $f=e^{-\beta u}-1$ by its linear approximation $f\simeq-\beta u$ is justified only when $|\beta u|\ll1$ , i.e. in the weak-interaction or high-temperature regime; in the repulsive core the linearization $-\beta u\to-\infty$ is catastrophically wrong, while the true $f\to-1$ is finite and integrable. The whole reason Mayer's trick works for hard cores — where a power series in $u$ diverges term by term — is that $f$ resums the exponential and stays bounded. Never linearize $f$ across the core.

Worked examples

1 · The Mayer function of a Lennard-Jones pair (confidence builder)

We evaluate $f(r)$ at the three landmark separations of the Lennard-Jones potential for an argon-like pair ( $\varepsilon/k_B=120$ K) at $T=300$ K, to see the sign structure explicitly.

Step 1. The reduced temperature is $\beta\varepsilon=(\varepsilon/k_B)/T = 120/300 = 0.40$ . At $r=\sigma$ the Lennard-Jones potential is $u=4\varepsilon[(1)^{12}-(1)^6]=0$ , so $f(\sigma)=e^{0}-1=\mathbf{0}$ . The Mayer bond vanishes exactly at the separation where the interaction potential changes sign, $u(r)=0$ .

Step 2. At the minimum $r_{\min}=2^{1/6}\sigma=1.122\,\sigma$ the potential is $u=-\varepsilon$ , so $f(r_{\min})=e^{\beta\varepsilon}-1=e^{0.40}-1=1.492-1=\mathbf{0.492}$ . The positive value indicates that configurations at the bottom of the attractive well are weighted about $49\%$ more strongly than in the non-interacting case.

Step 3. Inside the core, at $r=0.8\,\sigma$ , the potential is $u/\varepsilon=4[(1/0.8)^{12}-(1/0.8)^6]=4[14.55-3.815]=+42.9$ , so $\beta u=17.2$ and $f=e^{-17.2}-1\approx\mathbf{-1.000}$ . The bond has saturated at $-1$ : the configuration is forbidden.

Step 4. So along a single radial line $f$ runs from $-1$ (core, forbidden) up through $0$ (at $r=\sigma$ ) to $+0.49$ (in the well) and back to $0$ (far apart) — bounded, sign-changing, and short-ranged, exactly the properties that make the cluster expansion converge. Contrast the raw Boltzmann factor $e^{-\beta u}$ , which runs from $0$ to $1.49$ back to $1$ and never becomes a small parameter.

2 · The second virial coefficient and the Boyle temperature (exam level)

We compute $B_2$ for the same Lennard-Jones argon at two temperatures to exhibit the sign flip, then estimate where $B_2$ crosses zero.

Step 1. The second virial coefficient is $B_2(T)=-2\pi\int_0^\infty(e^{-\beta u(r)}-1)r^2\,\mathrm{d}r$ . Numerically integrating for the Lennard-Jones potential at $T=1000$ K (so $\beta\varepsilon=0.12$ , weak attraction) gives $B_2=+0.83\,\sigma^3$ — positive, because the hard core's excluded volume dominates when the well is thermally shallow.

Step 2. At $T=100$ K ( $\beta\varepsilon=1.2$ , strong attraction) the same integral gives $B_2=-7.3\,\sigma^3$ — negative and large, because the deep positive bump of $f$ in the well now outweighs the core.

Step 3. Converting to laboratory units at $T=300$ K with $\sigma=0.34$ nm, $\int f\,\mathrm{d}^3r=5.5\times10^{-29}\,\mathrm{m^3}$ , so $B_2=-\tfrac12\int f\,\mathrm{d}^3r\times N_A = \mathbf{-16\ cm^3/mol}$ — in excellent agreement with the measured argon value ( $\approx-16$ cm³/mol at $300$ K).

Step 4. Since $B_2>0$ at $1000$ K and $B_2<0$ at $100$ K and $300$ K, it must cross zero somewhere in between: the Boyle temperature $T_B$ , where the gas behaves ideally to first order in density. For argon $T_B\approx 410$ K. The single sign-changing integral of the Mayer function has produced a measurable, temperature-dependent equation-of-state coefficient and predicted the Boyle temperature — an illustration of the central role of the Mayer $f$ -function: a single cluster integral predicts the leading temperature dependence of the equation of state.

Origins · 1 minLennard-Jones potential and how this was discovered

Lennard-Jones potential

the 12–6 model pair potential · the standard test bed for the Mayer function

The Lennard-Jones potential, $u(r)=4\varepsilon\big[(\sigma/r)^{12}-(\sigma/r)^{6}\big]$ , is the canonical pair potential on which the Mayer $f$ -function is illustrated and tested. Proposed by John Lennard-Jones in 1924, it captures the two essential features of neutral-atom interactions with just two parameters: a steep repulsive wall $\propto r^{-12}$ from the overlap of electron clouds (Pauli exclusion at the atomic scale), and a gentle attractive tail $\propto r^{-6}$ from the induced-dipole van der Waals dispersion force. The potential vanishes at $r=\sigma$ , reaches its minimum depth $-\varepsilon$ at $r_{\min}=2^{1/6}\sigma$ , and decays to zero at large separation. Feeding it into $f=e^{-\beta u}-1$ produces exactly the bounded, sign-changing bond of this article: $f=-1$ in the core, $f=e^{\beta\varepsilon}-1>0$ at the minimum, $f=0$ at $\sigma$ and at infinity. Integrating that bond gives second virial coefficients that reproduce measured gas non-ideality — for argon, $B_2\approx-16$ cm³/mol at room temperature and a Boyle temperature near $410$ K. The Lennard-Jones model remains the default interatomic potential in molecular simulation, and it is the workhorse example throughout imperfect-gas theory.

History

1873

— van der Waals's equation of state anticipates hard-core repulsion and attractive forces phenomenologically.

1924

— Lennard-Jones proposes the 12–6 pair potential, giving a two-parameter model of neutral-atom interaction.

1927

— Ursell introduces cluster functions, the mathematical seed of the graph expansion.

1937

— Mayer defines the

f

-function and the cluster expansion, turning the configuration integral into a graph sum.

1940

— Mayer and Mayer's Statistical Mechanics codifies the

f

-function and virial machinery.

Going further

Texts. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 10 (cluster expansion and virial coefficients). Kardar, Statistical Physics of Particles, §5.2. Hansen & McDonald, Theory of Simple Liquids. McQuarrie, Statistical Mechanics, Ch. 12. Hirschfelder, Curtiss & Bird, Molecular Theory of Gases and Liquids (the Lennard-Jones $B_2$ tables). The limiting behaviour of $f$ , the Lennard--Jones benchmark values, and the temperature dependence of $B_2$ —including the experimental argon value of approximately $-16\,\mathrm{cm^3\,mol^{-1}}$ at $300\,\mathrm{K}$ —all follow directly from the definition $f=e^{-\beta u}-1$ and the radial integral $B_2=-2\pi\int_0^\infty\!\left(e^{-\beta u(r)}-1\right)r^2\,dr$ .

Prerequisites. Configuration integral · Canonical and grand canonical ensembles · Pairwise interaction potentials (Lennard-Jones, hard sphere).

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

Why subtract exactly 1? Why not some other constant?

Because $1$ is the value the pair Boltzmann factor $e^{-\beta u}$ takes when nothing is happening — at large separation where $u\to0$ . Subtracting the "no-interaction" value is what makes the remainder $f$ vanish where the particles do not interact, and it is precisely that vanishing that gives $f$ finite range and lets products of $f$ localize into graphs. Subtract any other constant and $f$ would be nonzero out to infinity, destroying the short-range property that makes the whole cluster expansion converge. The $1$ is not arbitrary; it is the ideal-gas reference baked into the definition.

How can f be a small expansion parameter if it equals −1 in the core?

It is not small pointwise — it is $-1$ throughout the core. What is "small" is the density: a term in the graph expansion with $k$ bonds requires $k$ simultaneous close encounters, and the probability of that scales as $n^k$ . So the expansion is controlled by density (or the packing fraction), not by the magnitude of $f$ itself. The core value $f=-1$ is exactly what you want — it says "this configuration is completely excluded" — and the integral $\int f\,\mathrm{d}^3r$ over the core is just $-\tfrac43\pi d^3$ , a finite excluded volume. Boundedness, not smallness, is the key property.

Why does B₂ change sign with temperature?

Because $f$ has two competing regions: the hard core where $f=-1$ (temperature-independent, always contributing a positive excluded-volume piece to $B_2$ ), and the attractive well where $f=e^{\beta|u|}-1>0$ (contributing a negative piece to $B_2$ , and growing rapidly as $T$ falls because $\beta|u|$ grows). At high $T$ the well contribution is thermally shallow and the core wins, so $B_2>0$ . At low $T$ the well contribution explodes and dominates, so $B_2<0$ . They exactly cancel at the Boyle temperature $T_B$ , where the gas is ideal to first order in density. This sign flip is a direct, measurable fingerprint of the shape of $u(r)$ .

Is the Mayer function only defined for pairwise potentials?

The two-particle Mayer function $f_{ij}=e^{-\beta u(r_{ij})}-1$ presupposes a pair potential, and the graph expansion in $f$ -bonds relies on pairwise additivity $U=\sum_{i<j}u_{ij}$ . If genuine three-body forces are present (like the Axilrod–Teller triple-dipole term), you must introduce additional three-body Mayer functions and new diagram types with three-body vertices. For most closed-shell atoms the three-body contribution is a few percent, so the pairwise $f$ -function is an excellent starting point; the framework extends systematically when higher-body forces matter.

What is the relationship between f and the pair correlation function g(r)?

They are close cousins. At leading order in density the radial distribution function is $g(r)\approx e^{-\beta u(r)}=1+f(r)$ — so $f(r)$ is essentially the leading-order deviation of $g(r)$ from its ideal value of $1$ . At higher density $g(r)$ picks up additional structure (indirect correlations mediated by other particles), which in the diagrammatic language are exactly the higher connected clusters built from more $f$ -bonds. So $f$ is the elementary bond, and $g(r)$ is the full dressed pair correlation that resums infinitely many $f$ -bonds. The second virial coefficient can be written with either, but $f$ is the microscopic building block.

Does the f-function work for the ideal quantum gases too?

Not directly — $f=e^{-\beta u}-1$ is built from a classical pair potential $u(r)$ , so it describes interactions, not quantum statistics. However, there is a beautiful parallel: for an ideal quantum gas the exchange effects of Bose/Fermi statistics can be packaged into an effective "statistical potential" and a corresponding statistical Mayer function, producing the exchange virial coefficient $\mp\lambda^3/2^{5/2}$ with no real interaction at all (owned by aq-sm-quantum-corrections-classical-ideal-gas). So the $f$ -function formalism inspired the way we treat quantum exchange as an effective interaction, but the classical $f$ here is about genuine forces.