Real gases — the virial expansion and van der Waals

Intuition Beginner

An ideal gas treats its molecules as points that never touch and never attract. Real molecules are different. Each one blocks a small volume of space — other molecules cannot enter it — and each one pulls weakly on its neighbours. At high temperature and low density these effects are invisible and the ideal-gas law $pV=NkT$ is accurate. Squeeze the gas or cool it, and the two effects show up as measurable departures from ideal behaviour.

The cleanest way to describe the departure is the compressibility factor $Z=pV/(NkT)$, equal to one for an ideal gas. For a real gas $Z$ depends on density and temperature, and the curve of $Z$ against pressure is a fingerprint of the fluid. The virial expansion writes $Z$ as one plus a sequence of density corrections. Each correction carries a coefficient $B_2(T),B_3(T),\dots$ that the microscopic interactions fix.

The van der Waals equation is the simplest model that captures both effects at once. It replaces the volume $V$ by $V-Nb$ to subtract the space the molecules occupy, and it lowers the pressure by $N^{2}a/V^2$ to model the attraction. The two parameters $a$ and $b$ are fit to data. Below a critical temperature the model predicts a liquid-vapor transition, and the same reduced equation then describes every simple fluid once $a$ and $b$ are scaled away.

Visual Beginner

A compressibility-factor chart. On the horizontal axis the reduced pressure, on the vertical axis $Z=pV/(NkT)$. Several isotherms (constant-temperature curves) are drawn. Above the Boyle temperature every isotherm rises above $Z=1$ as pressure grows, because repulsion dominates. Below the Boyle temperature each isotherm first dips below $Z=1$, because attraction dominates, then climbs back through $Z=1$ and above as the molecules are forced together.

The virial expansion is the Taylor-like series for each isotherm in powers of density around $Z=1$. The second virial coefficient $B_2(T)$ is its first slope: negative when attraction wins, positive when repulsion wins, and zero at the Boyle temperature.

Worked example Beginner

Take carbon dioxide, whose critical temperature $T_c=304.1$ K sits just above room temperature — the reason supercritical CO2 is a practical industrial solvent. The measured critical point is $p_c=72.8$ atm and molar volume $V_c=94.0$ cm³ per mole. These three numbers anchor the van der Waals fit.

Step 1 — fit the two parameters from $T_c$ and $p_c$, the most accurately measured constants. The relations $b=R{T}_c/(8p_c)$ and $a=27R^2{T}_c^2/(64p_c)$ with $R=8.314$ J/(mol K) give $b=4.28\times 10^{-5}$ m³/mol and $a=0.366$ Pa m⁶/mol². These are the excluded volume and the attraction strength per mole squared.

Step 2 — predict the critical volume from $V_c=3b=128$ cm³/mol. The measured value is $94$ cm³/mol. The model overshoots by about 36 percent, the same systematic error for every simple fluid.

Step 3 — check the critical compressibility factor $Z_c=p_c{V}_c/(R{T}_c)$. The model forces $Z_c=3/8=0.375$. The measured value for CO2 is $0.274$. The gap is the signature failure of mean-field theory at the critical point.

Step 4 — predict the second virial coefficient at $300$ K from $B_2=b-a/(RT)$. With $a/(RT)=147$ cm³/mol and $b=42.8$ cm³/mol, the model gives $B_2=-104$ cm³/mol. The measured value is about $-122$ cm³/mol: attraction dominates, and the model lands within 15 percent.

The lesson: a two-parameter fit captures the qualitative shape of the whole liquid-vapor transition. Quantitatively it is rough near the critical point and accurate enough far from it. The virial expansion tightens the quantitative picture order by order.

Check your understanding Beginner

Formal definition Intermediate+

This unit complements the sibling treatment in 08.11.03, which derives the virial coefficients from the grand-canonical cluster expansion and resums them into the van der Waals equation. Here the emphasis is dual: the virial equation of state as a phenomenological series in the compressibility factor, and the equation of state as reconstructed from the pair correlation function $g(r)$ through two independent exact routes that must agree.

The compressibility factor of a fluid is the dimensionless ratio

$$Z=\frac{pV}{N{k}_{B}T}=\frac{p}{\rho k_{B}T},\rho =\frac{N}{V},$$

equal to one for an ideal gas. Kamerlingh Onnes (1901) proposed ^{[Kamerlingh Onnes 1901]} a systematic density expansion of $Z$ — the virial equation of state:

$$Z=1+B_2(T)\rho +B_3(T){\rho }^2+B_4(T){\rho }^3+\cdots =1+\sum _{n\ge 2}{B}_n(T){\rho }^{n-1},$$

with the virial coefficients $B_n(T)$ functions of temperature alone. The series converges for sufficiently low density and develops a singularity at condensation. The Boyle temperature $T_B$ is defined by $B_2(T_B)=0$; there $Z\to 1$ to first order and the gas behaves ideally across a wide density range.

For a classical gas with pairwise potential $u(r)$ the configurational integral is

$$Q_N(V,T)={\int }_{V^N}\exp (-\beta \sum _{i<j}u(r_{ij}))d^{3N}r={\int }_{V^N}\prod _{i<j}(1+f(r_{ij}))d^{3N}r,$$

with the Mayer $f$-function $f(r)=e^{-\beta u(r)}-1$. Expanding the product generates all labelled graphs on $\{1,\dots ,N\}$. Ursell (1927) ^{[Ursell 1927]} first showed, and Mayer (1937) ^{[Mayer 1937]} reframed graph-theoretically, that $\log Q_N$ organises by connected-graph topology into a density series whose coefficients are the cluster integrals

$$b_n(T)=\frac{1}{n!V}{\int }_{V^n}\left(\text{sum of connected n-vertex f-bond graphs}\right)d^{3n}r.$$

The virial coefficients are polynomials in the $b_n$; equivalently each $B_n$ is the sum of all irreducible (biconnected) cluster integrals on $n$ vertices. The first is

$$B_2(T)=-\frac{1}{2}{\int }_{{\mathbb{R}}^3}f(r)d^{3}r=-2\pi {\int }_0^{\infty }(e^{-\beta u(r)}-1)r^{2}dr.$$

The pair correlation function $g(r)$ measures, relative to random, the probability of finding a second molecule at distance $r$ from a tagged one:

$$\rho g(r)=\text{average local number density at distance }r\text{ from a tagged particle}.$$

It is directly measurable by X-ray and neutron scattering: $g(r)$ is the Fourier transform of the static structure factor $S(k)$. The equation of state can be reconstructed from $g(r)$ by either of two exact routes that must agree for the exact $g(r)$.

• Virial (pressure) route. Scaling the container and differentiating the configurational integral gives

$$Z=1-\frac{2\pi \rho }{3k_{B}T}{\int }_0^{\infty }\frac{du}{dr}g(r)r^{3}dr.$$

• Compressibility route. The isothermal compressibility ${\kappa }_T=-(1/V)(\partial V/\partial p{)}_T$ satisfies the compressibility equation

$$\rho k_{B}T{\kappa }_T=1+\rho {\int }_{{\mathbb{R}}^3}(g(r)-1)d^{3}r\equiv S(0),$$

where $S(0)$ is the long-wavelength structure factor; integrating $(\partial \rho /\partial p{)}_T={\kappa }_T{\rho }^{-1}$ with respect to $\rho$ recovers a second expression for $Z$.

The Ornstein-Zernike equation ^{[Ornstein-Zernike 1914]} splits the total correlation $h(r)=g(r)-1$ into a direct correlation $c(r)$ and an indirect chain,

$$h(r)=c(r)+\rho \int c(|\mathbf{r}-{\mathbf{r}}^{\prime }|)h(r^{\prime })d^3{r}^{\prime },$$

a convolution equation whose Fourier form $S(k)=1/[1-\rho \hat{c}(k)]$ encodes the critical divergence $S(0)\to \infty$ when $\hat{c}(0)\to 1/\rho$.

Van der Waals equation (phenomenological). Rather than derive $Z(\rho ,T)$ from a specified $u(r)$, van der Waals (1873) ^{[van der Waals 1873]} proposed the two-parameter closed form

$$(p+\frac{a}{V_m^2})(V_m-b)=RT,$$

with $V_m=V/N_A$ the molar volume, $a>0$ an attraction parameter, and $b>0$ an excluded-volume parameter. Expanding for large $V_m$ recovers the virial series with $B_2(T)=b-a/(RT)$ and $B_n=b^{n-1}$ for $n\ge 3$, so the model compresses the entire interaction into two constants. The critical point satisfies $\partial p/\partial V={\partial }^{2}p/\partial V^2=0$, giving $T_c=8a/(27Rb)$, $V_c=3b$ (per mole), $p_c=a/(27b^2)$, and $Z_c=3/8$.

Law of corresponding states. In reduced variables $\tilde{T}=T/T_c$, $\tilde{p}=p/p_c$, $\tilde{V}=V/V_c$, the van der Waals equation collapses to

$$(\tilde{p}+\frac{3}{{\tilde{V}}^2})(\tilde{V}-\frac{1}{3})=\frac{8}{3}\tilde{T},$$

with no microscopic parameters remaining. Guggenheim (1945) ^{[Guggenheim 1945]} showed that the reduced coexistence curves of Ne, Ar, Kr, Xe, N2, O2, CO, and CH4 collapse onto a single universal curve to within a few percent — the empirical content of the corresponding-states principle.

Common slips

• $B_2(T)$ is temperature-dependent and changes sign at the Boyle temperature. Treating it as a fixed constant of the gas is wrong.
• The virial and compressibility routes give identical results only for the exact $g(r)$. Approximate integral-equation closures (Percus-Yevick, hypernetted-chain) break the equality, and the size of the disagreement is a standard diagnostic of closure quality.
• $Z_c=3/8=0.375$ is a universal van der Waals prediction. Real simple fluids cluster near $0.27$ to $0.29$. The discrepancy is not measurement error but the failure of mean-field theory at the critical point.
• The density virial expansion does not converge through the two-phase region. Below $T_c$ the gas-liquid coexistence must be inserted (Maxwell construction) before any virial description applies to the homogeneous branches.

Key derivation Intermediate+

Theorem (virial equation of state). For a classical fluid with pair potential $u(r)$ and pair correlation function $g(r)$, the compressibility factor is

$$Z=1-\frac{2\pi \rho }{3k_{B}T}{\int }_0^{\infty }\frac{du}{dr}g(r)r^{3}dr.$$

Derivation. From $F=-k_{B}T\log Z_N$ with $Z_N=Q_N/(N!{\lambda }_T^{3N})$, the pressure is $p=-(\partial F/\partial V{)}_{N,T}=k_{B}T(\partial \log Q_N/\partial V)$. Scale the container isotropically by $\mathbf{r}\mapsto \lambda \mathbf{r}$, so $V\mapsto {\lambda }^{3}V$ at fixed $\{N,T\}$, while the integration variables are held in their scaled positions. Then $r_{ij}\mapsto \lambda r_{ij}$ and the Jacobian contributes a factor ${\lambda }^{3N}$ that cancels against the rescaled domain, leaving

$$\frac{\partial \log Q_N}{\partial V}{|}_V=\frac{N}{V}-\frac{\beta }{3V}⟨\sum _{i<j}{r}_{ij}{u}^{\prime }(r_{ij})⟩.$$

The first term is ideal; the second is the ensemble average of the virial of the pair force. By the definition of $g(r)$, the pair average of any radial function $\varphi (r)$ is $⟨{\sum }_{i<j}\varphi (r_{ij})⟩=(N\rho /2)\int \varphi (r)g(r)d^{3}r$. Setting $\varphi (r)=r{u}^{\prime }(r)$ and using $d^{3}r=4\pi r^{2}dr$ converts the sum into the stated integral, and dividing by $p=\rho k_{B}TZ$ yields the formula. $\square$

Theorem (mean-field critical exponents of the van der Waals equation). Write the reduced van der Waals equation in deviations $v=(V-V_c)/V_c$ and $t=(T_c-T)/T_c$ from the critical point. To leading order

$$\pi \equiv \tilde{p}-1=-4t+6tv-\frac{3}{2}{v}^3+O(v^4,t{v}^2).$$

This single Landau-form expansion yields the four mean-field critical exponents $\alpha =0$, $\beta =1/2$, $\gamma =1$, $\delta =3$.

Derivation. Set $\tilde{v}=1+v$, $\tilde{p}=1+\pi$, $\tilde{t}=1-t$ in $(\tilde{p}+3/{\tilde{v}}^2)(\tilde{v}-1/3)=(8/3)\tilde{t}$ and expand, keeping terms up to combined order three in $\{t,v\}$. The linear terms cancel by the critical-point conditions $\partial p/\partial V={\partial }^{2}p/\partial V^2=0$; the quadratic terms cancel by the same conditions; what remains is the stated cubic. Each exponent is read off as follows.

• Critical isotherm ($\delta$). At $t=0$, $\pi =-(3/2)v^3$, so $|\pi |\propto |v{|}^3$ and $\delta =3$.
• Order parameter ($\beta$). Below $T_c$ ($t>0$) the gas-liquid coexistence densities are the two non-zero roots of $\partial \pi /\partial v=0$ at the coexistence pressure, equivalently where the odd part $6tv-(3/2)v^3$ vanishes: $v=\pm 2\sqrt{t}$. The density gap ${\rho }_L-{\rho }_G\propto 4\sqrt{t}$, giving $\beta =1/2$.
• Compressibility ($\gamma$). Along the critical isochore $v=0$, $\partial \pi /\partial v=6t$, so ${\kappa }_T\propto (\partial p/\partial V{)}^{-1}\propto |t{|}^{-1}$, giving $\gamma =1$.
• Specific heat ($\alpha$). In the van der Waals model the molar internal energy is $U_m=(3/2)RT-a/V_m$; along the critical isochore $V_m=V_c$ it is linear in $T$ with no singularity, so $C_V$ stays finite (a jump, not a divergence): $\alpha =0$. $\square$

Bridge. The two routes to the equation of state — the virial theorem applied to $g(r)$ and the compressibility integral of $g-1$ — build toward the whole machinery of liquid-state theory, and the requirement that they agree is the foundational reason that approximate closures are tested for thermodynamic consistency. The virial expansion itself appears again in 08.01.01 (the partition function is the generating object whose logarithm the cluster expansion computes) and in 08.02.01 (mean-field theory, where the same resummation produces the Curie-Weiss equation of the magnet). The central insight is that $B_2(T)=b-a/(RT)$ is the meeting point: a microscopic two-body integral on one side and the leading term of the phenomenological van der Waals fit on the other. The bridge is that any short-ranged interaction admits this dual description — graph-resummed (Ursell-Mayer) on the microscopic side, parameter-fit (van der Waals, Redlich-Kwong, Peng-Robinson) on the macroscopic side — and putting these together identifies the van der Waals equation as the Landau mean-field theory of the liquid-gas transition whose critical exponents 08.05.01 generalise beyond mean field via the Wilson renormalisation group.

Exercises Intermediate+

Advanced results Master

Theorem (two pressure routes and thermodynamic consistency). For a classical fluid the compressibility factor $Z$ admits two integral representations in terms of the pair correlation function: the virial route $Z=1-(2\pi \rho /3k_{B}T)\int u^{\prime }(r)g(r)r^{3}dr$ and the compressibility route $Z=1+{\int }_0^{\rho }[\int (g_{\rho }-1)d^{3}r]d{\rho }^{\prime }$. For the exact $g(r)$ the two routes coincide; for any approximate closure they differ, and the residual $\Delta Z=Z_{\text{vir}}-Z_{\text{comp}}$ is the standard diagnostic of closure quality.

The equivalence is not a coincidence: both routes are rearrangements of the same configurational integral $Q_N$. The virial route differentiates $Q_N$ with respect to volume at fixed particle number; the compressibility route differentiates twice with respect to chemical potential in the grand-canonical ensemble (where number fluctuates). Each manipulation is exact, so the exact $g(r)$ satisfies both. Integral-equation theories (Percus-Yevick, hypernetted-chain, Rogers-Young) replace the exact $g(r)$ by an approximate solution of the Ornstein-Zernike equation plus a closure, and the closure breaks one of the two identities. The Rogers-Young closure repairs this by mixing the Percus-Yevick and hypernetted-chain closures with a parameter tuned to set $\Delta Z=0$.

Theorem (hard-sphere virial series and Carnahan-Starling). For hard spheres of diameter $\sigma$ the virial coefficients are temperature-independent pure numbers times ${\sigma }^{3(n-1)}$: $B_2=(2\pi /3){\sigma }^3$, $B_3=(5{\pi }^2/18){\sigma }^6$, $B_4=(1{\pi }^3-28{\pi }^2/3+8\pi \sqrt{3}){\sigma }^9/12$, and so on. The Carnahan-Starling Pade resummation ^{[Carnahan-Starling 1969]}

$$Z=\frac{1+\eta +{\eta }^2-{\eta }^3}{(1-\eta {)}^3},\eta =\frac{\pi \rho {\sigma }^3}{6},$$

recovers the exact $B_2,B_3,B_4$ and matches simulation data for the hard-sphere fluid to within the percent level up to the freezing transition at $\eta \approx 0.49$.

The hard-sphere fluid is the reference system of liquid-state physics: its equation of state is known from simulation to high precision, and the Carnahan-Starling formula captures it with a closed-form rational function. Real fluids are then treated as a hard-sphere reference plus a perturbative attractive tail (Barker-Henderson, Weeks-Chandler-Andersen theories), a separation that supersedes the van der Waals excluded-volume term with a quantitatively accurate repulsive backbone. The van der Waals form $V\to V-Nb$ is the crudest possible approximation to the Carnahan-Starling result, valid only at low packing fraction where $Z\approx 1+4\eta +\cdots$ and $b=(2\pi /3){\sigma }^3$.

Theorem (mean-field universality of the liquid-gas transition). The four van der Waals critical exponents $\alpha =0$, $\beta =1/2$, $\gamma =1$, $\delta =3$ are the Landau saddle-point values of the Ising universality class. They are exact above the upper critical dimension $d_c=4$ and are corrected below $d_c=4$ by the Wilson-Fisher fixed point to the three-dimensional values $\alpha \approx 0.110$, $\beta \approx 0.326$, $\gamma \approx 1.237$, $\delta \approx 4.8$.

The identification of the liquid-gas transition with the Ising universality class is structural: both have a scalar order parameter (density deviation for the fluid, magnetisation for the magnet), a ${\mathbb{Z}}_2$ symmetry (particle-hole in the lattice-gas formulation of the fluid, spin-flip in the magnet), and a Landau free energy of the same quartic form. The van der Waals equation is the saddle-point evaluation of that Landau functional, and its exponents are the mean-field values. Fluctuations renormalise them below $d_c=4$, so the experimental exponents of CO2, argon, and every other simple fluid match the Ising values rather than the van der Waals values. The corresponding-states law survives because it concerns the shape of the coexistence curve in reduced variables, which is universal within the Ising class, while the exponents quantify its singularity at $T_c$.

Theorem (Joule-Thomson inversion from $B_2(T)$). The Joule-Thomson coefficient ${\mu }_{JT}=(\partial T/\partial p{)}_H$ of a real gas at low pressure is determined by the second virial coefficient: ${\mu }_{JT}=(1/C_p)(Td{B}_2/dT-B_2)$. The inversion temperature $T_{\text{inv}}$ at which ${\mu }_{JT}=0$ (the boundary between cooling and heating on expansion) satisfies $Td{B}_2/dT=B_2$.

The Joule-Thomson effect — the temperature change of a gas on throttled expansion — is the engineering signature of non-ideality, and its low-pressure behaviour is governed entirely by $B_2(T)$. For the van der Waals form $B_2=b-a/(RT)$, the inversion temperature is $T_{\text{inv}}=\sqrt{a/(Rb)}\cdot \sqrt{2}$... more directly, $Td{B}_2/dT=a/(RT)=B_2$ at inversion gives $2a/(RT)=b$, so $T_{\text{inv}}=2a/(Rb)=2T_B$. Every van der Waals gas inverts at twice its Boyle temperature. For CO2 this predicts $T_{\text{inv}}\approx 2050$ K, far above room temperature — which is why CO2 cools on expansion at $300$ K and is the working fluid of most refrigeration cycles. Hydrogen and helium, with very low $a$, have $T_{\text{inv}}$ below room temperature and heat up on expansion at $300$ K, a classic exception that any liquefier design must respect.

Synthesis. The virial expansion is the foundational reason that the macroscopic equation of state of any simple fluid is reconstructible from the pair potential and the pair correlation function, and the van der Waals equation is exactly the leading two-term resummation of that expansion into a closed phenomenological form. The central insight is that the two exact routes — the virial theorem applied to $g(r)$ and the compressibility integral of $g-1$ — are dual to each other and must agree for the exact $g(r)$, with their disagreement under approximate closures the standard diagnostic of liquid-state theory. Putting these together, the same density-expansion structure generalises to the mean-field critical exponents $\alpha =0,\beta =1/2,\gamma =1,\delta =3$, which are the Landau saddle-point values that the Wilson-Fisher renormalisation group 08.05.01 corrects below the upper critical dimension $d_c=4$. The bridge is that the liquid-gas transition, the Curie-Weiss magnet 08.02.01, and the ${\varphi }^4$ field theory are three presentations of one mean-field structure whose virial expansion is the Mayer-Ursell cluster series and whose phenomenological closure is the van der Waals equation; the same structure appears again in 08.01.01, where the partition function is the generating object from which all of these expansions descend.

Full proof set Master

Proposition (compressibility equation from number fluctuations). In the grand canonical ensemble the isothermal compressibility satisfies

$$\rho k_{B}T{\kappa }_T=\frac{⟨N^2⟩-⟨N{⟩}^2}{⟨N⟩}=1+\rho {\int }_{{\mathbb{R}}^3}(g(r)-1)d^{3}r.$$

Proof. In the grand canonical ensemble $\Xi ={\sum }_N{z}^N{Z}_N$ with $z=e^{\beta \mu }/{\lambda }_T^3$. The mean particle number is $⟨N⟩=z\partial \log \Xi /\partial z$ and its variance is $⟨N^2⟩-⟨N{⟩}^2=z\partial ⟨N⟩/\partial z=k_{B}T(\partial ⟨N⟩/\partial \mu {)}_T$. Since $\Omega =-pV=-k_{B}T\log \Xi$ and $⟨N⟩=-(\partial \Omega /\partial \mu {)}_{T,V}$, differentiating again gives the variance identity. Now $(\partial ⟨N⟩/\partial \mu {)}_{T,V}=(\partial ⟨N⟩/\partial p{)}_{T,V}\cdot (\partial p/\partial \mu {)}_{T,V}$. The Gibbs-Duhem relation at fixed $T$ gives $dp=\rho d\mu$, so $(\partial p/\partial \mu {)}_T=\rho$. Substituting: $⟨N^2⟩-⟨N{⟩}^2=k_{B}T\rho (\partial ⟨N⟩/\partial p{)}_{T,V}$. Dividing by $⟨N⟩$ and using ${\kappa }_T=(1/V)(\partial V/\partial p{)}_{T,⟨N⟩}=(1/\rho )(\partial \rho /\partial p{)}_T$ together with $(\partial ⟨N⟩/\partial p{)}_{T,V}=V\rho {\kappa }_T$ yields the first equality. For the second, write the pair count $⟨{\sum }_{i\ne j}\varphi (r_{ij})⟩={\rho }^{2}V\int \varphi (r)g(r)d^{3}r$ with $\varphi =1$, and subtract the ideal (Poisson) contribution $⟨N⟩$ to obtain $⟨N^2⟩-⟨N{⟩}^2=⟨N⟩+\rho ⟨N⟩\int (g-1)d^{3}r$, which after division by $⟨N⟩$ gives the second equality. $\square$

Proposition (low-density consistency of the two routes). At low density the virial route and the compressibility route both yield $B_2(T)=-(1/2){\int }_{{\mathbb{R}}^3}f(r)d^{3}r$, with $f=e^{-\beta u}-1$.

Proof. The compressibility-route computation is Exercise 4. For the virial route, substitute the low-density form $g(r)=e^{-\beta u(r)}=1+f(r)$ into $Z=1-(2\pi \rho /3k_{B}T){\int }_0^{\infty }{u}^{\prime }(r)g(r)r^{3}dr$ and isolate the $O(\rho )$ correction. Integrate by parts in $r$: $-{\int }_0^{\infty }{u}^{\prime }(r)e^{-\beta u(r)}{r}^{3}dr=-[u{r}^3{e}^{-\beta u}{]}_0^{\infty }+{\int }_0^{\infty }u(r)(d/dr)(r^3{e}^{-\beta u})dr$. After inserting $f=e^{-\beta u}-1$ and using $(d/dr)e^{-\beta u}=-\beta u^{\prime }{e}^{-\beta u}$, the boundary terms vanish for a short-ranged potential and the surviving integral collapses to $-{\int }_0^{\infty }(e^{-\beta u}-1)(d/dr)(r^3)dr/(3)+(\beta /3)\int u{u}^{\prime }{e}^{-\beta u}{r}^{3}dr$. A second integration by parts on the thermal piece returns $-(1/3){\int }_0^{\infty }f(r)(3r^2)dr=-{\int }_0^{\infty }f(r)r^{2}dr$. Multiplying by $-2\pi \rho /(k_{B}T)$ and collecting: the $O(\rho )$ term is $-(2\pi \rho /3k_{B}T)\cdot 3{\int }_0^{\infty }{u}^{\prime }(1+f)r^{3}dr$; the $u^{\prime }\cdot 1$ piece is a boundary term that vanishes, and the $u^{\prime }\cdot f$ piece reduces by the above to $\rho {\int }_0^{\infty }f{r}^{2}dr$. Converting to three dimensions, $B_2=-(1/2)\int f{d}^{3}r=-2\pi {\int }_0^{\infty }f{r}^{2}dr$, matching the compressibility route and the cluster result. $\square$

Proposition (Ornstein-Zernike divergence at criticality). In the Ornstein-Zernike framework $h=c+\rho c\star h$, the structure factor is $S(k)=1/[1-\rho \hat{c}(k)]$. At the critical point $\hat{c}(0)\to 1/\rho$, so $S(0)\to \infty$ and ${\kappa }_T\to \infty$; this is the microscopic origin of the mean-field compressibility exponent $\gamma =1$.

Proof. Fourier-transform the Ornstein-Zernike equation: $\hat{h}(k)=\hat{c}(k)+\rho \hat{c}(k)\hat{h}(k)$, giving $\hat{h}(k)=\hat{c}(k)/[1-\rho \hat{c}(k)]$. The structure factor $S(k)=1+\rho \hat{h}(k)=1/[1-\rho \hat{c}(k)]$. The compressibility equation gives $\rho k_{B}T{\kappa }_T=S(0)=1/[1-\rho \hat{c}(0)]$. As $T\to T_c$ from above, ${\kappa }_T$ diverges, which requires the denominator to vanish: $\rho \hat{c}(0)\to 1$. Expanding $\hat{c}(0;T)=(1/\rho )-A(T-T_c)+\cdots$ gives $S(0)\sim 1/[A(T-T_c)]\propto (T-T_c{)}^{-1}$, hence ${\kappa }_T\propto (T-T_c{)}^{-1}$ and $\gamma =1$ at mean-field level. The renormalisation group modifies the relation $\hat{c}(0;T)\to 1/\rho$ to a non-analytic approach and replaces the exponent with the three-dimensional Ising value $\gamma \approx 1.237$. $\square$

Connections Master

• Partition function 08.01.01. The configurational integral $Q_N$ is the spatial half of the canonical partition function, and its logarithm is the object that the Ursell-Mayer expansion reorganises into the density series whose coefficients are the virial coefficients $B_n(T)$. Without the partition function as generating object, neither the cluster expansion nor the two $g(r)$ routes to pressure has a setting.

• Sibling unit on the cluster derivation 08.11.03. The sibling derives the virial coefficients from the grand-canonical Mayer expansion and resums them into the van der Waals equation, then pursues the Yang-Lee zero theorem and the Wilson renormalisation-group view. This unit supplies the complementary side: the phenomenological virial series in the compressibility factor, the pair-correlation-function reconstruction, and the full four-exponent mean-field critical analysis.

• Mean-field theory 08.02.01. The van der Waals equation is the mean-field equation of state of the liquid-gas transition in the precise sense that the attractive term $-a/V_m^2$ replaces the actual pair attraction by its density average. This is structurally the same move as the Weiss molecular field of ferromagnetism, and both share the upper-critical-dimension story $d_c=4$.

• Critical exponents 08.05.01. The mean-field exponents $\alpha =0,\beta =1/2,\gamma =1,\delta =3$ derived here are the Landau values that the critical-exponents unit corrects to the three-dimensional-Ising numbers via the Wilson-Fisher fixed point. The liquid-gas transition belongs to the Ising universality class.

• Correlation functions 08.05.02. The pair correlation function $g(r)$ that drives both pressure routes here is the same object whose long-distance behaviour defines the correlation length $\xi$ and its critical divergence $\xi \sim |T-T_c{|}^{-\nu }$. The Ornstein-Zernike tail of $h(r)=g(r)-1$ is the bridge between the equation of state and the correlation-function singularity.

• Free energy 08.01.04. Both pressure routes differentiate the Helmholtz free energy $F=-k_{B}T\log Z$, and the Maxwell construction that resolves the van der Waals loop below $T_c$ is the convexification of $F(V)$ in $V$ at fixed $T$, equivalently the Legendre transform to the Gibbs free energy $G(P,T)$ whose common-tangent construction sets the coexistence pressure.

Historical & philosophical context Master

Johannes Diderik van der Waals presented his equation in his 1873 Leiden thesis Over de Continuiteit van den Gas- en Vloeistoftoestand ^{[van der Waals 1873]}, introducing the excluded-volume parameter $b$ and the attractive-mean-field parameter $a$ and deriving the law of corresponding states from the universal reduced form of the equation. The work earned him the 1910 Nobel Prize and supplied the first molecular theory of the liquid-gas transition.

The phenomenological side was completed by Heike Kamerlingh Onnes, who in 1901 ^{[Kamerlingh Onnes 1901]} proposed the virial expansion $p{V}_m/RT=1+B_2/V_m+B_3/V_m^2+\cdots$ as a systematic, purely empirical description of real-gas isotherms fit from the precise measurements of the Leiden cryogenic laboratory. The virial coefficients became the standard language in which experimental equation-of-state data are reported to this day.

The microscopic foundation arrived in two steps. Harold Ursell in 1927 ^{[Ursell 1927]} gave the first expansion of the configurational integral in powers of density using connected-cluster functions, and Joseph Mayer in 1937 ^{[Mayer 1937]} reframed it graph-theoretically with the $f$-function $f=e^{-\beta u}-1$, showing that each virial coefficient is a sum over irreducible cluster integrals. The Ornstein-Zernike split of correlations into direct and indirect pieces ^{[Ornstein-Zernike 1914]} then turned the compressibility route into a workable computational tool, and its long-range $1/r$ tail at $T_c$ explained critical opalescence. Guggenheim's 1945 corresponding-states compilation ^{[Guggenheim 1945]} confirmed that the reduced coexistence curves of eight simple fluids collapse onto a single universal curve — the empirical anchor of the universality hypothesis that the renormalisation group later explained. The Carnahan-Starling resummation ^{[Carnahan-Starling 1969]} closed the loop for the hard-sphere reference system, giving liquid-state theory the accurate repulsive backbone that the van der Waals excluded-volume term only crudely approximates.

Bibliography Master

@phdthesis{vanderWaals1873,
  author = {van der Waals, Johannes Diderik},
  title  = {Over de Continu{\"i}teit van den Gas- en Vloeistoftoestand},
  school = {Universiteit Leiden},
  year   = {1873}
}

@article{KamerlinghOnnes1901,
  author  = {Kamerlingh Onnes, Heike},
  title   = {Expression of the Equation of State of Gases and Liquids by Means of Series},
  journal = {Communications from the Physical Laboratory of the University of Leiden},
  number  = {71},
  year    = {1901}
}

@article{Ursell1927,
  author  = {Ursell, Harold D.},
  title   = {The Evaluation of {G}ibbs' Configuration Integral},
  journal = {Proc. Cambridge Phil. Soc.},
  volume  = {23},
  year    = {1927},
  pages   = {685--697}
}

@article{Mayer1937,
  author  = {Mayer, Joseph E.},
  title   = {Statistical Mechanics of Condensing Systems. {I}},
  journal = {J. Chem. Phys.},
  volume  = {5},
  year    = {1937},
  pages   = {67--73}
}

@article{OrnsteinZernike1914,
  author  = {Ornstein, Leonard S. and Zernike, Frits},
  title   = {Accidental Deviations of Density and Opalescence at the Critical Point of a Single Substance},
  journal = {Proc. K. Akad. Wet. Amsterdam},
  volume  = {17},
  year    = {1914},
  pages   = {793--806}
}

@article{Guggenheim1945,
  author  = {Guggenheim, Edward A.},
  title   = {The Principle of Corresponding States},
  journal = {J. Chem. Phys.},
  volume  = {13},
  year    = {1945},
  pages   = {253--261}
}

@article{CarnahanStarling1969,
  author  = {Carnahan, Norman F. and Starling, Kenneth E.},
  title   = {Equation of State for Nonattracting Rigid Spheres},
  journal = {J. Chem. Phys.},
  volume  = {51},
  year    = {1969},
  pages   = {635--636}
}

@book{MayerMayer1940,
  author    = {Mayer, Joseph E. and Mayer, Maria Goeppert},
  title     = {Statistical Mechanics},
  publisher = {John Wiley \& Sons},
  year      = {1940}
}

@book{HansenMcDonald2013,
  author    = {Hansen, Jean-Pierre and McDonald, Ian R.},
  title     = {Theory of Simple Liquids},
  edition   = {4},
  publisher = {Academic Press},
  year      = {2013}
}

@book{PathriaBeale2011,
  author    = {Pathria, R. K. and Beale, Paul D.},
  title     = {Statistical Mechanics},
  edition   = {3},
  publisher = {Elsevier},
  year      = {2011}
}

@book{McQuarrie2000,
  author    = {McQuarrie, Donald A.},
  title     = {Statistical Mechanics},
  publisher = {University Science Books},
  year      = {2000}
}

@book{Kubo1965,
  author    = {Kubo, Ryogo},
  title     = {Statistical Mechanics},
  publisher = {North-Holland},
  year      = {1965}
}