Third and Higher Virial Coefficients

The second virial coefficient captures the leading pairwise correction to the ideal gas, but it is only the first term in a systematic density expansion. The higher virial coefficients $B_3,B_4,\dots$ encode the cumulative effect of three, four, and more particles being simultaneously close. A tempting mistake is to imagine that $B_n$ demands genuine $n$ -body forces; in fact, for a pairwise-additive potential the higher coefficients are built entirely from products of pair Mayer functions $f_{ij}=e^{-\beta u(r_{ij})}-1$ . The correlations are many-body even though the forces are not — a distinction this article insists on. The central technical result, inherited from the cluster machinery, is that the virial coefficients are governed not by all connected clusters but only by the irreducible (star, biconnected) ones. When one inverts the pressure–fugacity series to reach the density expansion, every diagram with an articulation point — a vertex whose removal disconnects the graph — cancels. This article owns the third coefficient in full: $B_3=-\tfrac13\int f_{12}f_{13}f_{23}\,\mathrm{d}\mathbf r_{12}\mathrm{d}\mathbf r_{13}$ , the single triangle, and the celebrated hard-sphere result $B_3=\tfrac58 B_2^2$ , which we compute from the geometry of overlapping excluded shells and confirm by Monte Carlo. Counting the star diagrams is the combinatorial heart of the subject. There is exactly one star on three points (the triangle), but ten on four points and dozens on five, each with a symmetry weight. Beyond the triangle the integrals are high-dimensional and, for realistic potentials, evaluated numerically. Hard spheres are the tractable benchmark: $B_2$ and $B_3$ are exact in closed form, $B_4$ has a known exact analytic value, and $B_5$ onward are precise Monte-Carlo numbers. Physically the sequence $\{B_n\}$ sets the density range over which the virial series converges and encodes the approach to liquid–gas coexistence — the doorway to van der Waals and, eventually, the critical point.

Key takeaways

The virial expansion $P/(nk_BT)=1+B_2 n+B_3 n^2+\cdots$ organizes departures from ideality as a power series in number density $n=N/V$ .
Each virial coefficient is built from irreducible (star) Mayer diagrams; the third is the single triangle, $B_3=-\tfrac13\int f_{12}f_{13}f_{23}\,\mathrm{d}\mathbf r_{12}\mathrm{d}\mathbf r_{13}$ .
Reducible two-bond chains, though connected, carry an articulation point and cancel exactly in the density expansion — algebraically $B_3=4b_2^2-2b_3$ , and the $b_2^2$ chain part is removed.
For hard spheres $B_3=\tfrac58 B_2^2$ exactly (with $B_2=\tfrac23\pi\sigma^3=4v_0$ ); the geometric factor $\tfrac58<1$ comes from overlap of the three excluded shells.
$B_4^{\text{HS}}=0.28695\,B_2^3$ is known in exact closed form (Boltzmann/van Laar); $B_5$ and beyond are high-precision Monte-Carlo numbers.
Key idea: the higher virial coefficients are single irreducible (star) integrals — for pairwise forces the many-body content lives in the correlations, not the forces; $B_3$ is the triangle, and hard spheres give $B_3=\tfrac58 B_2^2$ .

The core insight

The higher virial coefficients are single irreducible (star) cluster integrals: $B_3=-\tfrac13\int f_{12}f_{13}f_{23}\,\mathrm{d}\mathbf r_{12}\mathrm{d}\mathbf r_{13}$ is the triangle, reducible chains cancel, and for hard spheres $B_3=\tfrac58 B_2^2$ .

Where this lives today

Higher virial coefficients are the backbone of accurate equations of state for real and model fluids. The hard-sphere coefficients $B_2$ through $B_{12}$ are the empirical input to the Carnahan–Starling equation of state, $Z=(1+\eta+\eta^2-\eta^3)/(1-\eta)^3$ in packing fraction $\eta=\tfrac\pi6 n\sigma^3$ , an extraordinarily accurate closed form built by recognizing the near-geometric decay of $B_{n+1}/B_2^n$ and resumming it. For real gases, $B_3(T)$ computed from ab initio pair and three-body potentials (helium especially) now underpins primary metrology — the same virial framework that fixes the Boltzmann constant through acoustic and dielectric gas thermometry. The subject also probes physics beyond pairwise additivity. Because $B_3$ for a pairwise potential is the triangle of pair bonds, measured deviations of $B_3$ from the pairwise prediction quantify genuine three-body forces such as the Axilrod–Teller triple-dipole interaction in the noble gases — the higher virial coefficients are a sensitive experimental window on many-body forces. And when the reduced coefficients stop decaying smoothly, the series is warning of condensation: the growth of $|B_n|$ sets the radius of convergence, and its breakdown near the critical point is the doorway to the renormalization-group treatment later in this track.

The physical picture

Picture the gas as a soup of transient bonds: a bond $f_{ij}$ is appreciable only when particles $i$ and $j$ are within interaction range. The cluster expansion catalogues every pattern of bonds, but the virial coefficients ask a sharper question than the raw cluster integrals do — not "how many ways can three particles be bonded?" but "how many ways can they be bonded so robustly that no single particle is a bottleneck?" A two-bond chain $1\!-\!2\!-\!3$ is connected, yet cutting vertex $2$ severs it: vertex $2$ is an articulation point, and the chain is reducible. The triangle $1\!-\!2\!-\!3\!-\!1$ has no such weak point — delete any vertex and it stays connected. The virial inversion's near-magical outcome is that all the chains and trees, which vastly outnumber the stars, cancel exactly, leaving only the rare irreducible diagrams. The first figure fixes this vocabulary at the three-particle level: the three reducible chains that cancel and the single irreducible triangle that survives to build $B_3$ . This is the thesis of the whole article on one plot.

Figure

Three-particle graphs: the three reducible two-bond chains (articulation point at the middle vertex, cancel in the density expansion) versus the single irreducible triangle (no cut vertex, the sole survivor that builds B₃)

The hard-sphere ratio $B_3=\tfrac58 B_2^2$ is the article's quantitative anchor. It says the three-body excluded-volume correction is not simply the square of the two-body one — naive multiplication would give $B_3=B_2^2$ . The reduction to $\tfrac58$ reflects that when particle $3$ must avoid both $1$ and $2$ , the two excluded shells overlap, so the combined exclusion is less than the product. The second figure plots the reduced hard-sphere coefficients $B_n/B_2^{\,n-1}$ : they fall off smoothly from $1$ , and the punchline is that the exact points sit below the naive "all equal to $1$ " line — the geometric overlap correction, made visible.

Figure

Reduced hard-sphere virial coefficients Bₙ/B₂ⁿ⁻¹ (exact/known values) falling below the naive Bₙ=B₂ⁿ⁻¹ line: B₂/B₂=1, B₃/B₂²=5/8=0.625, B₄/B₂³=0.287, B₅/B₂⁴=0.110 — the excluded-shell overlap reduces every higher coefficient

Higher coefficients matter most where the series is stressed: near condensation. The radius of convergence of $\sum_n B_n n^n$ is set by the growth of $|B_n|$ , and a singularity at the condensation density is encoded in their asymptotics — a deep link between the combinatorics of star graphs and the physics of the phase transition that this track later treats with the renormalization group.

The core equations

The virial expansion writes the compressibility factor as a density series; the first coefficient is $1$ (ideal), the rest measure interactions. This is the equation of state $B_n$ feed.

\frac{P}{n k_BT}=1+B_2(T)\,n+B_3(T)\,n^2+\cdots=\sum_{k=0}^{\infty}B_{k+1}(T)\,n^{k},\qquad B_1\equiv1 \qquad (1)

where: $P$ — pressure; $n=N/V$ — number density; $k_BT$ — thermal energy; $B_k(T)$ — the $k$ -th virial coefficient. The coefficient $B_2$ (owned by aq-sm-second-virial-coefficient) and the full inversion (owned by aq-sm-virial-expansion) are recalled by name; this article supplies $B_3$ and the structure of $B_{n\ge3}$ . Each virial coefficient is a single irreducible (star) cluster integral, tied to it by a clean rational prefactor. We use the rooted normalization of aq-sm-irreducible-cluster-integrals, in which $\beta_k$ carries a $1/k!$ and one pinned vertex.

B_{k+1}=-\frac{k}{k+1}\,\beta_k,\qquad \beta_k=\frac{1}{k!\,V}\int\Big(\sum_{\text{stars on }k+1}\ \prod_{(ij)}f_{ij}\Big)\,\mathrm{d}^{3(k+1)}r \qquad (2)

where: $\beta_k$ — the $k$ -th irreducible cluster integral over $k+1$ particles (rooted, $1/k!$ normalization); "stars on $k+1$ " — biconnected graphs with no articulation vertex. At $k=1$ , $\beta_1=\int f\,\mathrm{d}^3r$ (a single bond, trivially a star) and $B_2=-\tfrac12\beta_1=-\tfrac12\int f$ ; the map is proved in aq-sm-irreducible-cluster-integrals, recalled here by name. For the third coefficient the only star on three vertices is the triangle. Writing it with the pinned vertex at the origin gives the standard two-integration form.

B_3=-\frac{2}{3}\,\beta_2=-\frac13\int f(r_{12})\,f(r_{13})\,f(r_{23})\,\mathrm{d}^3r_{12}\,\mathrm{d}^3r_{13} \qquad (3)

where: $\beta_2=\tfrac12\int f_{12}f_{13}f_{23}\,\mathrm{d}^3r_{12}\mathrm{d}^3r_{13}$ is the rooted triangle integral (the $\tfrac12$ is the $1/2!$ ), so $B_3=-\tfrac23\beta_2=-\tfrac13\times(\text{full triangle }T)$ . The two conventions (rooted $\beta_2$ vs full triangle $T=2\beta_2$ ) give the same $B_3$ ; the warning below flags the notational trap. Comparing with the ordinary cluster integrals $b_l$ makes the reduction concrete: $B_2=-b_2$ , but $B_3\ne -b_3$ — the reducible chains inside $b_3$ must be removed.

B_2=-b_2,\qquad B_3=4b_2^2-2b_3 \qquad (4)

where: $b_l$ — the connected cluster integrals of aq-sm-cluster-integrals (which include the reducible chains). The combination $4b_2^2-2b_3$ is what the density inversion produces; substituting $b_3=\tfrac16(3J^2+T)$ with $J=\int f$ shows the $J^2$ (chain) part cancels, leaving $B_3=-\tfrac13 T$ — the pure triangle. For hard spheres the triangle integral is a geometry problem, and the answer is the famous rational $\tfrac58$ .

B_2^{\text{HS}}=\frac{2\pi}{3}\sigma^3=4v_0,\qquad B_3^{\text{HS}}=\frac{5}{8}\,B_2^2=\frac{5\pi^2}{18}\sigma^6 \qquad (5)

where: $\sigma$ — hard-sphere diameter; $v_0=\tfrac{\pi}{6}\sigma^3$ — the volume of one sphere; the excluded volume per particle is $B_2=4v_0$ . The reduction $\tfrac58$ (not $1$ ) is the overlap of the excluded shells, verified symbolically and by Monte Carlo. Beyond the triangle, exact closed forms are rare. The fourth hard-sphere coefficient is known analytically (Boltzmann 1899; van Laar); the fifth onward are Monte-Carlo numbers.

B_4^{\text{HS}}=\Big(\tfrac{2707}{4480}+\tfrac{219\sqrt2}{2240\,\pi}-\tfrac{4131}{4480\,\pi}\arccos\tfrac{1}{3}\Big)B_2^3=0.28695\,B_2^3,\qquad B_5^{\text{HS}}\approx0.110252\,B_2^4 \qquad (6)

where: the $\arccos\tfrac{1}{3}$ argument is essential — the exact closed form evaluates to $0.286950\,B_2^3$ . (The AllQFT base article wrote $\arccos\tfrac{1}{\sqrt3}$ , which gives $0.36785$ — a base error corrected here; see the warning.) $B_5^{\text{HS}}\approx0.110252\,B_2^4$ is a Ree–Hoover Monte-Carlo value; still higher $B_n$ are tabulated numerically.

Derivation AHard-sphere third virial coefficient: B₃ = ⅝ B₂² from excluded-shell overlap

The goal is to evaluate the triangle integral $B_3=-\tfrac13\int f_{12}f_{13}f_{23}\,\mathrm{d}^3r_{12}\mathrm{d}^3r_{13}$ for hard spheres and obtain the exact ratio $B_3/B_2^2=\tfrac58$ . The strategy is to note that for hard spheres each Mayer bond is simply $-1$ inside contact and $0$ outside, so the triangle integral counts (with a sign) the phase-space volume in which all three particles mutually overlap. Fixing one particle at the origin, the integral over the other two is the volume of configurations where $r_{12}<\sigma$ , $r_{13}<\sigma$ , and $r_{23}<\sigma$ simultaneously — a geometric overlap of excluded regions whose reduction below the naive $B_2^2$ is the whole point.

Step 1. The hard-sphere Mayer function. For hard spheres of diameter $\sigma$ , $u=+\infty$ for $r<\sigma$ and $0$ for $r>\sigma$ , so $f(r)=e^{-\beta u}-1=-1$ for $r<\sigma$ and $0$ for $r>\sigma$ . Why: inside contact the Boltzmann factor vanishes, giving $f=-1$ ; outside there is no interaction, so $f=0$ . The bond is a pure indicator of overlap.

Step 2. Reduce to the second coefficient first. $B_2=-\tfrac12\int f\,\mathrm{d}^3r=-\tfrac12(-1)\!\int_0^\sigma 4\pi r^2\mathrm{d}r=\tfrac12\cdot\tfrac43\pi\sigma^3=\tfrac23\pi\sigma^3=4v_0$ . Why: the single-bond integral is just the volume of the exclusion sphere of radius $\sigma$ ; halving it and flipping the sign gives the positive excluded-volume $B_2$ , four times a molecule's volume.

Step 3. Set up the triangle. $B_3=-\tfrac13\,T$ with $T=\int f_{12}f_{13}f_{23}\,\mathrm{d}^3r_{12}\mathrm{d}^3r_{13}$ . With particle $1$ at the origin, $f_{12}f_{13}f_{23}=(-1)^3=-1$ exactly when $|\mathbf r_2|<\sigma$ , $|\mathbf r_3|<\sigma$ , and $|\mathbf r_2-\mathbf r_3|<\sigma$ , and $0$ otherwise. So $T=-\,\mathrm{Vol}\{(\mathbf r_2,\mathbf r_3):\text{all three pairwise distances}<\sigma\}$ . Why: the product of three $f$ 's is $(-1)^3=-1$ only in the triple-overlap region; elsewhere at least one bond is zero. The triangle integral is minus that six-dimensional overlap volume.

Step 4. Evaluate the overlap. Integrating $\mathbf r_3$ first for fixed $\mathbf r_2$ gives the lens volume of two spheres of radius $\sigma$ whose centers are $r_{12}$ apart, $V_{\text{lens}}(r_{12})=\tfrac{\pi}{12}(2\sigma+r_{12})(\sigma-r_{12})^2\cdot 4$ for $r_{12}<\sigma$ ; integrating this against $\mathbf r_2$ over $r_{12}<\sigma$ yields $T=-\tfrac{15}{8}B_2^2$ . Why: the inner integral is the standard two-sphere intersection (lens) volume; the outer radial integral of $V_{\text{lens}}(r)r^2$ over $[0,\sigma]$ produces the rational multiple $\tfrac{15}{8}B_2^2$ of the squared exclusion volume — the overlap reduces it from the naive $2B_2^2$ .

Step 5. Apply the prefactor. $B_3=-\tfrac13 T=-\tfrac13\big(-\tfrac{15}{8}B_2^2\big)=\tfrac{5}{8}B_2^2=\tfrac{5}{8}\big(\tfrac23\pi\sigma^3\big)^2=\tfrac{5\pi^2}{18}\sigma^6$ . Why: the $-\tfrac13$ from the irreducible map turns the overlap volume into the third virial coefficient; the result is positive (net repulsion) and smaller than $B_2^2$ because the excluded shells overlap. The clean $\tfrac58$ is a pure geometric number, independent of $\sigma$ . *Verified: the triangle integral $T$ , the map $B_3=-\tfrac13 T$ , and the exact ratio $B_3/B_2^2=\tfrac58=0.625$ are checked symbolically and by importance-sampled Monte Carlo (giving $0.62499$ ) in verify_higher-virial-coefficients.py, matching Hansen & McDonald Ch. 3.*

Go deeper · Derivation B · 1 minThe reducible chains cancel: B₃ = 4b₂² − 2b₃ leaves only the triangle

Derivation BThe reducible chains cancel: B₃ = 4b₂² − 2b₃ leaves only the triangle

The goal is to show explicitly that the two-bond chain diagrams, which are connected and enter the cluster integral $b_3$ , cancel in the equation of state, so that $B_3$ is the irreducible triangle alone. The strategy is to work to third order in fugacity, invert the density series to express $z(n)$ , substitute into the pressure series, and read off the $n^3$ coefficient — then decompose $b_3$ into its chain and triangle parts and watch the chain contribution be cancelled by the $b_2^2$ term. This is the concrete, order-by-order demonstration of the topological reduction owned in full by aq-sm-irreducible-cluster-integrals.

Step 1. The two series. With $b_1=1$ , the pressure is $P/k_BT=z+b_2z^2+b_3z^3+\cdots$ and the density $n=z\,\partial_z(P/k_BT)=z+2b_2z^2+3b_3z^3+\cdots$ . Why: the grand-canonical pressure is the connected-cluster sum in fugacity; the density is its logarithmic derivative, which multiplies each term by its power — the different weightings drive the inversion.

Step 2. Invert the density series. Solving $n=z+2b_2z^2+3b_3z^3$ for $z$ by series reversion gives $z=n-2b_2n^2+(8b_2^2-3b_3)n^3+\cdots$ . Why: term-by-term Lagrange reversion; the quadratic coefficient flips sign and doubles, and the cubic collects both $b_2^2$ and $b_3$ — this is where the chains and the square get entangled.

Step 3. Substitute into the pressure. Inserting $z(n)$ into $P/k_BT=z+b_2z^2+b_3z^3$ and collecting powers gives $P/k_BT=n-b_2n^2+(4b_2^2-2b_3)n^3+\cdots$ , so $B_2=-b_2$ and $B_3=4b_2^2-2b_3$ . Why: the algebra is exact to cubic order; the $n^3$ coefficient is $4b_2^2-2b_3$ , not $-b_3$ , precisely because reducible structure has leaked in through $b_2^2$ .

Step 4. Decompose b₃ and cancel. Write the connected $b_3=\tfrac16(3J^2+T)$ with $J=\int f$ (the three chains give $3J^2$ , the triangle gives $T$ ) and $b_2=\tfrac12 J$ . Then $B_3=4(\tfrac12 J)^2-2\cdot\tfrac16(3J^2+T)=J^2-(J^2+\tfrac13 T)=-\tfrac13 T$ . Why: the chain part $3J^2$ inside $b_3$ is exactly cancelled by $4b_2^2=J^2$ , so only the irreducible triangle $T$ survives — the reducible diagrams have disappeared from the equation of state, confirming $B_3=-\tfrac13 T=-\tfrac23\beta_2$ . *Verified: the reversion $z(n)$ , the identity $B_3=4b_2^2-2b_3$ , and the cancellation $4b_2^2-2b_3=-\tfrac13 T$ with $b_3=\tfrac16(3J^2+T)$ are all checked symbolically in verify_higher-virial-coefficients.py, matching Kardar SPP §5.*

Limiting cases

The higher coefficients simplify or are exactly known in several regimes.

Case	Result	Meaning
Hard spheres, $B_3$	$B_3=\tfrac58 B_2^2$ , $B_2=\tfrac23\pi\sigma^3$	Exact closed form; the single triangle over hard-sphere overlap volumes.
Hard spheres, $B_4$	$B_4=0.28695\,B_2^3$	Exact analytic value (Boltzmann/van Laar), with $\arccos\tfrac13$ ; $B_5\approx0.110252\,B_2^4$ (Monte Carlo).
High temperature	$B_n(T)\to B_n^{\text{hard core}}$	The attractive tail is negligible; each $B_n$ approaches its repulsive excluded-volume value.
Low density	$P/(nk_BT)=1+B_2 n+\mathcal O(n^2)$	Higher $B_n$ enter only at higher powers; truncation is controlled by $n\sigma^3$ .
Weak coupling	$\beta u\ll1:\ B_3\sim(\beta u)^3$	Each bond $f\approx-\beta u$ carries one power; the triangle (three bonds) scales as $(\beta u)^3$ .

Common trapTwo traps: β₂ conventions and the B₄ arccosine

Two notational hazards live here. First, the symbol $\beta_2$ is defined two ways in the literature: some texts (and the AllQFT base of the neighbouring article) call the full triangle integral $T=\int f_{12}f_{13}f_{23}$ the irreducible integral and write $B_3=-\tfrac13 T$ ; others use the rooted $\beta_2=\tfrac12 T$ (a $1/2!$ from pinning one vertex) and write $B_3=-\tfrac23\beta_2$ . Both give the identical $B_3=-\tfrac13 T$ — the difference is only in what " $\beta_2$ " names. This article uses the rooted convention throughout, consistent with aq-sm-irreducible-cluster-integrals. Second, the exact hard-sphere $B_4$ closed form contains $\arccos\tfrac13$ , not $\arccos\tfrac{1}{\sqrt3}$ : the former evaluates to $0.286950\,B_2^3$ (the accepted value), the latter to $0.36785$ — an error present in the base export and corrected here. Never conflate the two, and never quote three-body integrals as three-body forces: for pairwise potentials $B_3$ is a product of pair bonds.

Worked examples

1 · The hard-sphere third virial coefficient in SI numbers (confidence builder)

We compute $B_2$ and $B_3$ for argon-sized hard spheres ( $\sigma=0.34$ nm) and express $B_3$ in cm³²/mol², the units of tabulated data.

Step 1. The exclusion sphere has radius $\sigma=0.34\times10^{-9}$ m, so $B_2=\tfrac23\pi\sigma^3=\tfrac23\pi(0.34\times10^{-9})^3=\tfrac23\pi(3.93\times10^{-29})=8.23\times10^{-29}\ \mathrm{m^3}$ per pair.

Step 2. In molar units, $B_2=N_A\,b_{\text{pair}}=6.022\times10^{23}\times8.23\times10^{-29}\ \mathrm{m^3/mol}=4.96\times10^{-5}\ \mathrm{m^3/mol}=\mathbf{49.6\ cm^3/mol}$ , the excluded-volume estimate (real argon $B_2$ turns negative at low $T$ from attraction, but the pure hard core gives this positive number).

Step 3. The third coefficient is $B_3=\tfrac58 B_2^2$ . Working per pair squared, $B_3=\tfrac58(8.23\times10^{-29})^2=\tfrac58(6.77\times10^{-57})=4.23\times10^{-57}\ \mathrm{m^6}$ .

Step 4. In molar units $B_3=N_A^2\times4.23\times10^{-57}=(6.022\times10^{23})^2\times4.23\times10^{-57}=3.63\times10^{47}\times4.23\times10^{-57}=1.53\times10^{-9}\ \mathrm{m^6/mol^2}=\mathbf{1530\ cm^6/mol^2}$ . A single geometric factor $\tfrac58$ has fixed the entire three-body correction from the two-body number.

2 · How much smaller is B₃ than the naive B₂²? (exam level)

We quantify the excluded-shell overlap by comparing the exact hard-sphere $B_3$ with the naive $B_2^2$ that would follow from treating the two exclusions as independent.

Step 1. The naive (independent-exclusion) estimate is $B_3^{\text{naive}}=B_2^2$ , obtained by squaring the two-body excluded volume as if particle $3$ 's two constraints never overlapped.

Step 2. The exact result is $B_3=\tfrac58 B_2^2=0.625\,B_2^2$ , from the triangle integral. The ratio of exact to naive is $\tfrac58=0.625$ .

Step 3. So the true three-body coefficient is $\mathbf{37.5\%\ smaller}$ than the naive product ( $1-\tfrac58=\tfrac38$ ). Physically, when particle $3$ is close enough to both $1$ and $2$ , those two exclusion zones overlap, so the combined forbidden volume is less than the sum — double counting is removed by the geometry.

Step 4. For the fourth coefficient the overlap is even more severe: $B_4=0.28695\,B_2^3$ versus a naive $B_2^3$ , a reduction to about $\mathbf{29\%}$ . The trend $1,\ 0.625,\ 0.287,\ 0.110,\dots$ of the reduced coefficients is the smooth decay plotted above, and it is exactly what makes the Carnahan–Starling closed form such an accurate resummation of the hard-sphere series.

Origins · 1 minClausius virial and how this was discovered

Clausius virial

Clausius's 1870 virial theorem, the historical root of the "virial" name (Latin vis, force) · the virial name and the mechanical origin of the pressure corrections

The word "virial" — from the Latin vis, viris (force) — was coined by Rudolf Clausius in 1870 for the quantity $-\tfrac12\sum_i \mathbf r_i\cdot\mathbf F_i$ , the "virial of the forces," in his theorem relating the time-averaged kinetic energy of a bounded mechanical system to the virial of the forces acting on it. For an ideal gas the virial is purely the wall (pressure) contribution and the theorem reproduces $PV=Nk_BT$ ; interparticle forces add an internal virial that shifts the pressure, and it was Kamerlingh Onnes who, in 1901, organized these shifts into the empirical power series $P/(nk_BT)=1+B_2 n+B_3 n^2+\cdots$ whose coefficients he named the virial coefficients. The higher coefficients $B_3,B_4,\dots$ are the systematic, order-by-order content of Clausius's internal virial: each measures how simultaneous close encounters of three, four, or more molecules correct the pressure. The modern route, through Mayer's cluster expansion, expresses them as irreducible (star) integrals over products of pair Mayer functions — a diagrammatic reincarnation of Clausius's mechanical idea that the deviations from ideality are the accumulated virial of the intermolecular forces.

History

1662

— Boyle establishes the ideal

PV=

const law, the baseline the virial coefficients later correct.

1870

— Clausius states the virial theorem and coins "virial" for the force sum controlling the pressure.

1899

— Boltzmann computes the exact hard-sphere fourth virial coefficient (later re-derived by van Laar).

1901

— Kamerlingh Onnes writes the empirical virial expansion and names the virial coefficients.

1937

— Mayer's cluster expansion expresses the coefficients as irreducible star integrals.

1964

— Ree and Hoover compute high-order hard-sphere virial coefficients by Monte Carlo.

Going further

Texts. Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 10 (cluster expansion and virial coefficients). Kardar, Statistical Physics of Particles, §5 (cumulant/cluster expansions). Hansen & McDonald, Theory of Simple Liquids, 4th ed., Ch. 3 (hard-sphere virial coefficients, Carnahan–Starling). Mayer & Mayer, Statistical Mechanics (1940), the original irreducible-cluster theory. Ree & Hoover, J. Chem. Phys. (1964), high-order hard-sphere coefficients.

Provenance. Base article: AllQFT track `sm`, "Third and Higher Virial Coefficients" (STAT.INTER.BNVIR.1248); sources Pathria & Beale Ch. 10, Kardar §5, Hansen & McDonald Ch. 3. All numeric and symbolic claims re-verified in verify_higher-virial-coefficients.py: the inversion $B_3=4b_2^2-2b_3$ , the chain cancellation to $-\tfrac13 T$ , the hard-sphere $B_2=4v_0$ and $B_3=\tfrac58 B_2^2$ (Monte Carlo $0.62499$ ), and the exact $B_4/B_2^3=0.286950$ . Base error corrected: the base wrote the $B_4$ closed form with $\arccos\tfrac{1}{\sqrt3}$ (evaluating to $0.36785$ ); the correct argument is $\arccos\tfrac13$ (evaluating to $0.286950$ , the accepted Boltzmann/van Laar value) — fixed here and flagged in-article. The base's two derivations used $\beta_2=$ full triangle with $B_3=-\tfrac13\beta_2$ ; this article uses the rooted $\beta_2=\tfrac12 T$ , $B_3=-\tfrac23\beta_2$ , consistent with aq-sm-irreducible-cluster-integrals, and notes the equivalence.

Prerequisites. Second virial coefficient · Virial expansion of the equation of state · Irreducible cluster integrals and the virial series.

Leads to. Van der Waals from the virial expansion · Maxwell construction and coexistence · Law of corresponding states.

Why do the reducible chain diagrams cancel — where do they actually go?

They cancel against the square of the second cluster integral when you convert from the fugacity series to the density series. The connected three-cluster $b_3$ contains both the triangle and three two-bond chains. When you invert $n(z)$ to $z(n)$ and substitute into $P(z)$ , the pressure's $n^3$ coefficient is $4b_2^2-2b_3$ , and the $4b_2^2$ piece — which is $J^2$ with $J=\int f$ — exactly cancels the $3J^2$ chain part hidden inside $2b_3=\tfrac13(3J^2+T)$ . Nothing is thrown away by hand; the density inversion does the bookkeeping. What remains is $-\tfrac13 T$ , the pure irreducible triangle. The chains reappear, in effect, as the reducible dressing of lower-order coefficients rather than as new physics at order $n^3$ .

Is B₃ ever negative, or is it always positive like the hard-sphere value?

For hard spheres $B_3=\tfrac58 B_2^2>0$ always — pure repulsion cannot make the three-body correction negative. But real potentials have attraction, and $B_3(T)$ does go negative at low temperature for the Lennard-Jones and square-well fluids: when the attractive well is deep compared with $k_BT$ , configurations where three molecules cluster in each other's wells lower the pressure, driving $B_3<0$ . As temperature rises, $B_3$ passes through zero and approaches its positive hard-core value. So the sign of $B_3$ , like that of $B_2$ , is a thermometer of the competition between the attractive tail and the repulsive core — but the crossover temperatures for $B_2$ and $B_3$ differ.

Why can't we just compute all the B_n exactly like B₂ and B₃?

Because the star integrals become high-dimensional and geometrically intricate. $B_2$ is a one-dimensional radial integral; $B_3$ is a single triangle, still doable in closed form for simple potentials. But $B_4$ already involves ten distinct star diagrams on four vertices (only for hard spheres does a closed form exist, and even that took from Boltzmann to the twentieth century to pin down), and $B_5$ and beyond have dozens of stars whose multidimensional integrals lack closed forms. They are evaluated by high-precision Monte-Carlo integration (Ree–Hoover and successors). This is why the practical strategy is not to sum the exact series but to resum it — the Carnahan–Starling equation packages the known hard-sphere coefficients into one accurate closed form.

If the forces are pairwise, how can B₃ be a "three-body" quantity at all?

The distinction is between forces and correlations. For a pairwise-additive potential the forces involve only two particles at a time, and $B_3$ contains no three-body force — it is literally $\int f_{12}f_{13}f_{23}$ , a product of pair bonds. But the correlation it measures is genuinely three-body: it counts configurations in which all three particles are simultaneously within range of one another, an event that no product of independent pair events reproduces (that is exactly why the overlap reduces $B_3$ below $B_2^2$ ). So "three-body" refers to the simultaneous three-particle correlation, not to a three-body force. Real three-body forces (like Axilrod–Teller) would add a separate contribution on top, and comparing measured $B_3$ with the pairwise prediction is how experiment detects them.

What does the ⅝ factor have to do with the shape of the excluded region?

Everything. If particle $3$ 's exclusion by particle $1$ and its exclusion by particle $2$ were independent events, the combined excluded phase space would be the product, giving $B_3=B_2^2$ . But when $1$ and $2$ are themselves within a diameter of each other (which they must be, since $f_{12}\ne0$ requires $r_{12}<\sigma$ ), their exclusion spheres for particle $3$ overlap heavily. The triangle integral computes exactly this overlapping (lens-shaped) volume, and doing the geometry gives the rational factor $\tfrac58$ : the true three-body excluded volume is $37.5\%$ smaller than the naive product. The number is purely geometric — independent of $\sigma$ , temperature, or any material parameter — which is why it is the same for every hard-sphere fluid.

How do higher virial coefficients signal the approach to condensation?

The virial series $\sum_n B_n n^{n-1}$ has a radius of convergence set by the large- $n$ growth of $|B_n|$ : if $|B_n|^{1/n}\to1/n_c$ , the series converges only for density below $n_c$ . A singularity of the equation of state at $n_c$ — physically, the onset of the liquid–gas instability — is therefore encoded in the asymptotic behaviour of the coefficients. As you approach condensation, successive $B_n$ stop decaying smoothly and the truncated series becomes unreliable; the divergence is not a failure of the method but a diagnostic of the phase transition. Extracting $n_c$ from the coefficient asymptotics is one classic (if delicate) way to estimate where a model fluid condenses, linking the star-graph combinatorics directly to the phase diagram treated in the following chapters.