Pair Correlation Function and the Equation of State

A crystal wears its structure on its sleeve: atoms sit on a lattice, and a single unit cell tells you everything. A liquid hides its structure. There is no lattice, no long-range order, and yet a liquid is emphatically not a random gas — stand on any molecule and the neighbours arrange themselves into shells, a crowded first ring, a quieter gap, a fainter second ring, fading with distance. The pair correlation function $g(r)$ is the instrument that reads this hidden order. It reports, for a particle sitting at the origin, how the local density of other particles varies with distance $r$ , normalized so that far away, where correlations die, $g(r)\to 1$ . What makes $g(r)$ the master key of liquid-state theory is a remarkable economy. For a system whose particles interact through a pairwise potential $u(r)$ , the entire equilibrium thermodynamics is fixed by $g(r)$ together with the density and temperature. No higher distribution function is needed, because every thermodynamic quantity is an average of a pair quantity, and the probability of finding a pair at separation $r$ is exactly $n^2 g(r)/2$ per unit volume. From this single function flow three independent exact relations — the energy equation, the pressure (virial) equation, and the compressibility equation — each an integral of $g(r)$ against the potential, its derivative, or nothing at all. This article derives all three routes, shows that the pressure route reproduces the virial expansion and its second coefficient $B_2$ when $g(r)$ is expanded in density, and connects $g(r)$ to the static structure factor $S(k)$ that X-ray and neutron scattering measure directly. It closes on a subtle and instructive fact: the three routes agree only when fed the exact $g(r)$ . Approximate closures make them disagree, and the size of that disagreement is a working diagnostic of how good the approximation is.

Key takeaways

The pair correlation function $g(r)$ is the normalized conditional density of particles at distance $r$ from a reference particle: $g(r)\to 0$ inside a hard core, peaks above 1 in the first shell, and $\to 1$ at large $r$ .
For a pairwise-additive potential, all thermodynamics follows from $g(r)$ through three exact routes — the energy, pressure (virial), and compressibility equations.
The pressure equation $\dfrac{P}{nk_BT}=1-\dfrac{n}{6k_BT}\displaystyle\int r\,u'(r)\,g(r)\,4\pi r^2\,\mathrm{d}r$ reduces to the virial series with $B_2=-\tfrac12\int f\,\mathrm{d}^3r$ when $g(r)\to e^{-\beta u}$ .
The compressibility equation $n k_BT\kappa_T=1+n\int[g(r)-1]\,\mathrm{d}^3r=S(0)$ uses no potential at all — it ties structure to density fluctuations and diverges at the critical point.
$g(r)$ and the structure factor $S(k)$ are a Fourier pair; forward scattering $S(0)$ measures the compressibility, and its divergence is critical opalescence.
The three routes agree only for the exact $g(r)$ ; for approximate closures they disagree — the "thermodynamic inconsistency" that grades an approximation.
Key idea: for pairwise interactions the configurational thermodynamics is entirely carried by the two-particle distribution; integrating $g(r)$ against $u(r)$ , against $r\,u'(r)$ , or over space alone yields the internal energy, the pressure, and the compressibility respectively.

The core insight

For pairwise interactions the configurational thermodynamics is entirely carried by the two-particle distribution; $g(r)$ is its dimensionless reduced form, and integrating it against $u(r)$ , against $r\,u'(r)$ , or simply over space yields the internal energy, the pressure, and the isothermal compressibility respectively.

Where this lives today

The pair correlation function is the workhorse of liquid-state physics and soft matter. X-ray and neutron diffraction measure $S(k)$ for everything from liquid metals to molten salts to water; inverting the Fourier relation returns $g(r)$ , whose first-peak position gives the nearest-neighbour distance, whose peak area gives the coordination number, and whose $k\to0$ limit gives the compressibility. In colloid science the same $g(r)$ and $S(k)$ are measured by light scattering and confocal microscopy, where the "atoms" are micron-sized spheres one can watch individually. Molecular-dynamics and Monte Carlo simulations routinely compute $g(r)$ by histogramming pair separations. From the resulting $g(r)$ , one may evaluate the energy, virial, or compressibility routes and compare them as a stringent consistency check on the simulation or on approximate liquid-state theories. Its reach extends well beyond simple fluids. For hard spheres the Carnahan–Starling equation, built as a consistency-weighted blend of the two Percus–Yevick routes, is accurate to better than a percent up to freezing and anchors the perturbation theories (Barker–Henderson, Weeks–Chandler–Andersen) that predict the equations of state of argon, methane and more complex molecular liquids. Near critical points, the Ornstein–Zernike tail of $g(r)-1$ and the divergence of $S(0)$ connect this classical machinery to the modern theory of critical phenomena and the renormalization group. Wherever matter is disordered but correlated — liquids, glasses, colloids, plasmas — $g(r)$ remains the first thing one measures and the first thing one computes.

The physical picture

Stand on one molecule of a dense liquid and look outward. Immediately around you is a correlation hole — no other particle can overlap your hard core, so $g(r)\approx 0$ there. Just beyond the core sits a crowded ring of nearest neighbours, the first coordination shell, where the local density rises well above average and $g(r)$ peaks at 2 or 3. Push the neighbours in, and the ones behind them are pushed out: a depleted trough follows, then a weaker second shell, then gentler oscillations that decay toward the uncorrelated value $g=1$ . The whole curve is a normalized snapshot of liquid architecture — the first-peak position gives the nearest-neighbour distance, its area gives the coordination number, and the decay length reports how far order propagates. The reason $g(r)$ suffices for thermodynamics is that thermodynamics asks only for averages of pair quantities. The total potential energy is a sum over pairs of $u(r_{ij})$ ; to average it you need only the probability of finding a pair at separation $r$ , which is precisely $n^2 g(r)/2$ per unit volume. Pressure and energy are therefore integrals of $g(r)$ against the potential — no three-body information is required for a two-body potential.

Figure

Pair correlation g(r): a dense liquid has a correlation hole (g→0 at small r), a first-shell peak (g>1, ~2.6), and decaying oscillations settling to 1 — contrasted with the flat ideal-gas g=1 and the low-density Boltzmann curve g(r)=e^{−βu}

The deepest route is the one that involves no potential at all. The integral $\int[g(r)-1]\,\mathrm{d}^3r$ counts the net excess or deficit of neighbours around a particle. That count is exactly the relative variance of the number of particles in a fixed volume — a fluctuation. Large fluctuations mean a long-ranged $g(r)-1$ and a large compressibility; near a critical point the correlations become infinitely long-ranged and the compressibility diverges. This is the structural origin of critical opalescence, the milky glow of a fluid poised at its critical point. Because $g(r)$ and the structure factor $S(k)$ are a Fourier-transform pair, the same function read in real space gives the shell structure and read in reciprocal space gives the scattering pattern that a neutron or X-ray beam actually records. Theory and experiment shake hands at $g(r)$ .

Figure

The structure factor S(k) is the Fourier partner of g(r): a first peak near k ≈ 2π/d (d = nearest-neighbour spacing) and the small-k limit S(0) = n k_B T κ_T equal to the compressibility route — which diverges at the critical point

The core equations

Everything begins with the two-particle density, the joint probability of finding one particle near $\mathbf r_1$ and another near $\mathbf r_2$ , obtained by integrating the Boltzmann weight over all the other coordinates. It is the raw object that $g(r)$ tames.

\rho^{(2)}(\mathbf r_1,\mathbf r_2)=N(N-1)\frac{\int e^{-\beta U}\,\mathrm{d}\mathbf r_3\cdots\mathrm{d}\mathbf r_N}{\int e^{-\beta U}\,\mathrm{d}\mathbf r_1\cdots\mathrm{d}\mathbf r_N} \qquad (1)

where: $\rho^{(2)}$ — two-particle density; $U=\sum_{i<j}u(r_{ij})$ — total configurational energy for a pairwise potential; $\beta=1/k_BT$ ; the factor $N(N-1)$ counts ordered pairs. In a homogeneous isotropic fluid $\rho^{(2)}$ depends only on the separation $r=|\mathbf r_1-\mathbf r_2|$ , and for an ideal gas it reduces to $n^2$ — particles are placed independently. Normalizing the two-particle density by its ideal-gas value $n^2$ isolates the interaction-induced structure and defines the radial distribution function, the dimensionless star of the show.

g(r)=\frac{1}{n^2}\,\rho^{(2)}(\mathbf r_1,\mathbf r_2),\qquad r=|\mathbf r_1-\mathbf r_2| \qquad (2)

where: $g(r)$ — pair correlation / radial distribution function (dimensionless); $n=N/V$ — number density. Reading it physically: $n\,g(r)\,4\pi r^2\,\mathrm{d}r$ is the average number of particles in a spherical shell of radius $r$ and thickness $\mathrm{d}r$ around any chosen particle, so $g(r)$ is that count relative to the uniform expectation $n\,4\pi r^2\,\mathrm{d}r$ . The internal energy is the first payoff. The excess energy beyond the ideal kinetic $\tfrac32 Nk_BT$ is simply the average pair energy, and the number of pairs at separation $r$ is set by $g(r)$ .

\frac{U}{N}=\frac{3}{2}k_BT+\frac{n}{2}\int u(r)\,g(r)\,4\pi r^2\,\mathrm{d}r \qquad (3)

where: the first term is the kinetic ideal-gas energy; the integral is the configurational (excess) energy per particle. The factor $n/2$ counts pairs without double-counting; for an ideal gas ( $u=0$ , or $g=1$ with $u=0$ ) the integral vanishes and $U/N=\tfrac32 k_BT$ as it must. The pressure follows from how the free energy responds to a change of volume, which rescales every separation. The result weights $g(r)$ by the virial $r\,u'(r)$ of the pair force.

\frac{P}{nk_BT} = 1-\frac{2\pi n}{3k_BT} \int_0^\infty r^3\frac{du(r)}{dr}\, g(r)\,dr \qquad (4)

where: $P$ — pressure; $r\,u'(r)$ — the virial of the pair force ( $-u'$ is the radial force); the leading $1$ is the ideal-gas term. Attractions ( $u'>0$ where the force pulls in) lower the pressure; a hard repulsive core ( $u'<0$ ) raises it — exactly the two competing effects that make a real gas depart from ideal. The compressibility route is the odd one out, and the most powerful: it contains no potential whatsoever. It ties the spatial integral of $g(r)-1$ to the isothermal compressibility through particle-number fluctuations.

n\,k_BT\,\kappa_T=1+n\int[g(r)-1]\,4\pi r^2\,\mathrm{d}r \qquad (5)

where: $\kappa_T=-\tfrac1V(\partial V/\partial P)_T=\tfrac1n(\partial n/\partial P)_T$ — isothermal compressibility; $h(r)\equiv g(r)-1$ — the total correlation function. Because no $u(r)$ appears, this route survives even when the potential is not pairwise additive — a crucial robustness the energy and pressure routes lack. Finally, $g(r)$ is not just a theorist's construct: its Fourier transform is the static structure factor that a scattering experiment measures, and the small- $k$ limit returns the compressibility.

S(k)=1+n\int[g(r)-1]\,e^{-i\mathbf k\cdot\mathbf r}\,\mathrm{d}^3r,\qquad S(0)=n\,k_BT\,\kappa_T \qquad (6)

where: $S(k)$ — static structure factor; $k$ — scattering wavevector. Setting $k=0$ makes the exponential unity and recovers Eq. (5): forward scattering is the compressibility, and its divergence near a critical point is critical opalescence made quantitative.

Route	Formula	Uses the potential?	Special virtue
Energy	$U/N=\tfrac32 k_BT+\tfrac{n}{2}\int u\,g\,4\pi r^2\mathrm{d}r$	yes ( $u$ )	direct thermal energy
Pressure (virial)	$\frac{P}{nk_BT} = 1-\frac{2\pi n}{3k_BT} \int_0^\infty r^3u'(r)\,g(r)\,dr$	yes ( $r u'$ )	gives the equation of state; reduces to $B_2$
Compressibility	$n k_BT\kappa_T=1+n\int(g-1)\,\mathrm{d}^3r=S(0)$	no	valid for non-pairwise systems; diverges at $T_c$

Derivation AThe pressure equation reduces to the second virial coefficient at low density

The goal is to show that the pressure (virial) equation is not an independent piece of physics but reduces to the virial expansion in the low-density limit: inserting the low-density limit $g(r)\to e^{-\beta u(r)}$ into Eq. (4) must reproduce the virial series $P/(nk_BT)=1+B_2\,n+\cdots$ with the standard second virial coefficient $B_2=-\tfrac12\int f\,\mathrm{d}^3r$ . The route is to substitute the limiting $g$ , rewrite the awkward $r\,u'\,e^{-\beta u}$ integrand as a total derivative of the Mayer function $f=e^{-\beta u}-1$ , and integrate by parts. The one thing to watch is the boundary term at infinity: it vanishes only because $f(r)$ is short-ranged, which is exactly the condition under which the whole cluster picture is valid.

Step 1. Insert the limiting g(r). At leading order in density $g(r)\to e^{-\beta u(r)}$ , so $\dfrac{P}{nk_BT}=1-\dfrac{n}{6k_BT}\displaystyle\int r\,u'(r)\,e^{-\beta u(r)}\,4\pi r^2\,\mathrm{d}r+\mathcal O(n^2)$ . Why: the higher-density corrections to $g$ generate $B_3,B_4,\dots$ ; keeping only the Boltzmann factor isolates the $B_2$ term.

Step 2. Turn the integrand into a total derivative. Note $\dfrac{\mathrm{d}}{\mathrm{d}r}e^{-\beta u}=-\beta u'\,e^{-\beta u}$ , so $u'\,e^{-\beta u}=-k_BT\,\dfrac{\mathrm{d}}{\mathrm{d}r}\!\left(e^{-\beta u}-1\right)=-k_BT\,f'(r)$ . Hence $r\,u'\,e^{-\beta u}=-k_BT\,r\,f'(r)$ . Why: writing the integrand through $f'$ lets the explicit $u'$ disappear and the $k_BT$ cancel the $\beta$ hiding in the temperature prefactor — the algebra is verified symbolically.

Step 3. Integrate by parts. The integral becomes $-\dfrac{n}{6k_BT}\displaystyle\int(-k_BT)\,r f'(r)\,4\pi r^2\,\mathrm{d}r=\dfrac{4\pi n}{6}\displaystyle\int r^3 f'(r)\,\mathrm{d}r$ , and $\displaystyle\int_0^\infty r^3 f'\,\mathrm{d}r=\big[r^3 f\big]_0^\infty-3\displaystyle\int_0^\infty r^2 f\,\mathrm{d}r=-3\displaystyle\int r^2 f\,\mathrm{d}r$ . Why: the boundary term $r^3 f$ vanishes at both ends — at $r=0$ because of the $r^3$ , at $r=\infty$ because $f$ decays faster than any power for a short-range potential.

Step 4. Collect and read off B₂. Assembling, $\dfrac{P}{nk_BT}=1-\dfrac{4\pi n}{6}\cdot 3\displaystyle\int r^2 f\,\mathrm{d}r=1-\tfrac12 n\displaystyle\int f(r)\,4\pi r^2\,\mathrm{d}r=1+B_2\,n$ , so $B_2=-\tfrac12\displaystyle\int f\,\mathrm{d}^3r$ . Why: this is exactly the second virial coefficient of the cluster expansion — the pressure route and the virial series agree, and for a hard sphere of diameter $\sigma$ ( $f=-1$ inside) it gives $B_2=\tfrac23\pi\sigma^3=4v_{\text{molecule}}$ . *The key steps — the integrand identity $r u' e^{-\beta u}=-k_BT r f'$ , the integration by parts $\int r^3 f'=-3\int r^2 f$ , the prefactor chain to $B_2=-\tfrac12\int f\,\mathrm{d}^3r$ , the hard-sphere result $B_2=\tfrac23\pi\sigma^3=4v_{\text{mol}}$ , and the numeric value $B_2\approx-16$ cm³/mol for Lennard-Jones argon at 300 K — all hold consistently.*

Go deeper · Derivation B · 1 minThe compressibility equation from grand-canonical number fluctuations

Derivation BThe compressibility equation from grand-canonical number fluctuations

Here the goal is to derive the compressibility route, Eq. (5), by a path that touches no potential — tying the spatial integral of $g(r)-1$ to the isothermal compressibility purely through the statistics of particle-number fluctuations. The strategy is to work in the grand canonical ensemble where $N$ fluctuates, use the standard fluctuation–response identity relating $\langle(\Delta N)^2\rangle$ to $(\partial N/\partial\mu)$ , express the same fluctuation through the one- and two-particle densities, and equate. The payoff to watch for is why the potential drops out: the fluctuation of $N$ is a property of the distribution of particles, not of the forces that produced it, which is what makes this route uniquely robust.

Step 1. Fluctuation–response identity. In the grand ensemble $\langle(\Delta N)^2\rangle=k_BT\left(\dfrac{\partial\langle N\rangle}{\partial\mu}\right)_{T,V}=n\,k_BT\,N\,\kappa_T$ , where the last equality uses $(\partial n/\partial\mu)_T=n^2\kappa_T$ from the Gibbs–Duhem relation. Why: this is the general link between the variance of an extensive quantity and the susceptibility conjugate to its field — here $N$ and $\mu$ — with no reference to interactions.

Step 2. Counting particle pairs. Pair counting gives $\langle N(N-1)\rangle=\int_V \mathrm{d}\mathbf r_1\int_V \mathrm{d}\mathbf r_2\,\rho^{(2)}(\mathbf r_1,\mathbf r_2)$ . Since $\langle(\Delta N)^2\rangle=\langle N(N-1)\rangle+\langle N\rangle-\langle N\rangle^2$ , the number fluctuation can be expressed in terms of the pair correlation function after substituting $\rho^{(2)}(\mathbf r_1,\mathbf r_2)=n^2g(r)$ . Why: $\langle N^2\rangle$ splits into the diagonal self term (giving $\langle N\rangle$ ) and the off-diagonal pair term (the two-particle density); subtracting $\langle N\rangle^2=\iint n^2$ leaves the $\rho^{(2)}-n^2$ integrand.

Step 3. Reduce to g(r). Using $\rho^{(2)}=n^2 g(r)$ and translational invariance to perform one volume integral, $\dfrac1N\langle(\Delta N)^2\rangle=1+n\displaystyle\int[g(r)-1]\,\mathrm{d}^3r$ . Why: the double integral collapses to $V\int$ over the relative coordinate; dividing by $N=nV$ leaves the intensive integral of the total correlation function $h=g-1$ .

Step 4. Equate the two expressions. Setting Step 1 equal to Step 3 gives $n\,k_BT\,\kappa_T=1+n\displaystyle\int[g(r)-1]\,\mathrm{d}^3r=S(0)$ . Why: no $u(r)$ ever entered Steps 2–3 — the compressibility route is therefore valid even for genuinely many-body potentials, where the energy and pressure equations (which assume pairwise additivity) fail. The $k\to0$ limit of $S(k)$ is this same sum rule, the experimentally tested compressibility relation. *As a consistency check: for the ideal gas $g=1\Rightarrow\int(g-1)=0\Rightarrow n k_BT\kappa_T=1$ , i.e. $\kappa_T=1/(nk_BT)$ , and $S(0)=1$ — both in accord with the Hansen–McDonald compressibility sum rule.*

Limiting cases

The formalism is anchored by a few clean limits that every approximation must respect.

Case	Result	Meaning
Ideal gas	$g(r)=1$ for all $r$	No correlations; the pressure equation gives $P=nk_BT$ and the compressibility equation gives $nk_BT\kappa_T=1$ .
Low density	$g(r)\xrightarrow{\,n\to0\,}e^{-\beta u(r)}$	The Boltzmann factor of the pair potential; substituting into the virial route reproduces $B_2=-\tfrac12\int f\,\mathrm{d}^3r$ .
Weak coupling / large $r$	$g(r)-1\approx-\dfrac{u(r)}{k_BT}$	Correlations follow the potential's tail at high $T$ ; $g\to1$ as $r\to\infty$ .
Critical point	$\displaystyle\int[g-1]\,\mathrm{d}^3r\to\infty,\ \kappa_T\to\infty$	Ornstein–Zernike long-range tail $g(r)-1\sim e^{-r/\xi}/r$ with $\xi\to\infty$ ; $S(0)$ diverges — critical opalescence.
Hard spheres	$B_2=\tfrac23\pi\sigma^3=4v_{\text{mol}}$	Pure excluded volume; four times the volume of one sphere, temperature-independent.

Common trapThe three equations of state agree only for the exact g(r)

The energy, pressure, and compressibility equations are exact identities — but they yield the same equation of state only when you feed them the exact $g(r)$ . Insert an approximate $g(r)$ (Percus–Yevick, HNC, etc.) into the energy, virial, and compressibility equations and they generally yield different thermodynamic predictions. This is not a flaw in the formulas; it is a signature that no approximate two-body distribution satisfies the full hierarchy of exact relations at once. The magnitude of the discrepancy is a working measure of the closure's accuracy — and consistency-enforcing schemes deliberately tune a parameter to make the two routes agree. A second, separate limitation: $g(r)$ alone does not fix the three-body and higher correlations, so it cannot deliver the thermodynamics of a genuinely many-body potential through the energy or pressure routes.

Worked examples

1 · The second virial coefficient of argon from g(r) (confidence builder)

We compute $B_2$ for Lennard-Jones argon at $300$ K through the pressure route with $g(r)\to e^{-\beta u(r)}$ , and check it against experiment.

Step 1. Use the Lennard-Jones potential $u(r)=4\varepsilon\big[(\sigma/r)^{12}-(\sigma/r)^{6}\big]$ with $\varepsilon/k_B=120$ K and $\sigma=0.34$ nm. At $T=300$ K the reduced inverse temperature is $\beta\varepsilon=120/300=0.40$ .

Step 2. In the low-density limit $g(r)=e^{-\beta u(r)}$ , so the pressure route reduces (previous derivation) to $B_2=-\tfrac12\int f(r)\,4\pi r^2\,\mathrm{d}r$ with $f=e^{-\beta u}-1$ . Numerically integrating gives $\int f(r)\,d^3r=5.47\times10^{-29}\ {\rm m^3}$ , so $B_2=-\frac12\int f(r)\,d^3r =-2.73\times10^{-29}\ {\rm m^3}$ .

Step 3. Converting to molar units, $B_2=-2.73\times10^{-29}\times6.022\times10^{23}\times10^{6}=\mathbf{-16.5\ cm^3/mol}$ .

Step 4. The measured second virial coefficient of argon at $300$ K is about $-16$ cm³/mol — excellent agreement. The negative sign says attraction wins at room temperature: argon is slightly more compressible than an ideal gas. A single structural input, $g(r)=e^{-\beta u}$ , run through the pressure route, has reproduced real gas non-ideality.

2 · The compressibility of the ideal gas from the structure factor (exam level)

We confirm that the compressibility route and the structure factor reduce correctly for the ideal gas, and read off what forward scattering measures.

Step 1. For the ideal gas there are no correlations, so $g(r)=1$ everywhere and $h(r)=g(r)-1=0$ . The compressibility equation (5) becomes $nk_BT\kappa_T=1+n\int 0\,\mathrm{d}^3r=1$ , hence $\kappa_T=1/(nk_BT)$ .

Step 2. Check against thermodynamics: the ideal gas has $P=nk_BT$ , so $\big(\partial P/\partial n\big)_T=k_BT$ and $\kappa_T=\tfrac1n\big(\partial n/\partial P\big)_T=\tfrac{1}{nk_BT}$ . The two agree — the structural route recovers the textbook ideal compressibility.

Step 3. The structure factor at $k=0$ is $S(0)=1+n\int[g-1]\,\mathrm{d}^3r=1$ . So an ideal gas scatters with $S(0)=\mathbf{1}$ : uncorrelated density fluctuations of unit relative variance.

Step 4. In a real fluid near its critical point, $\int[g-1]\,\mathrm{d}^3r$ grows without bound as the correlation length $\xi\to\infty$ , so $S(0)\to\infty$ and $\kappa_T\to\infty$ . Forward scattering intensity, proportional to $S(0)$ , blows up — the fluid turns milky. That is critical opalescence, and it is nothing but the compressibility route seen with light.

Origins · 1 minRadial distribution function and how this was discovered

Radial distribution function

the pair correlation g(r) bridging microscopic structure and thermodynamics (Ornstein & Zernike 1914; Kirkwood, Yvon, Born & Green 1930s–40s; measured by X-ray and neutron scattering) · three thermodynamic routes from one function

The radial distribution function $g(r)$ is the central structural quantity of the modern theory of liquids: a dimensionless, density-normalized measure of how the presence of one particle biases the positions of the rest. Its thermodynamic power was recognized in stages. Ornstein and Zernike, working in 1914 on the puzzle of critical opalescence, introduced the total and direct correlation functions and the relation that still bears their names, showing how a diverging correlation length produces the runaway forward scattering that makes a critical fluid glow. Through the 1930s and 1940s Kirkwood, Yvon, Born and Green built the hierarchy of coupled equations linking $g(r)$ to higher distribution functions, and derived the three exact routes — energy, pressure, and compressibility — by which $g(r)$ determines the thermodynamics of a pairwise-interacting fluid. The compressibility route is special because it invokes no potential at all: it follows from number-fluctuation statistics and therefore survives even for many-body forces. The great practical lesson of the subject, learned from approximate closures such as Percus–Yevick and the hypernetted chain, is thermodynamic consistency: the three routes coincide only for the exact $g(r)$ , so their disagreement is a quantitative grade for any approximation. Because $g(r)$ is the Fourier partner of the measured structure factor $S(k)$ , it is the point where liquid-state theory meets X-ray and neutron diffraction — a rare quantity that is at once fundamental and directly observed.

History

1914

— Ornstein and Zernike introduce the direct correlation function and explain critical opalescence as a diverging correlation length.

1927

— Zernike and Prins interpret X-ray scattering from liquids in terms of the radial distribution function.

1935

— Kirkwood proposes the superposition approximation for three-body correlations.

1946

— Born, Green, Yvon and Bogoliubov develop the coupled hierarchy of distribution-function equations.

1958

— Percus and Yevick derive the closure that yields the first accurate analytic hard-sphere g(r).

1960

s — neutron and X-ray scattering routinely deliver experimental g(r) for real liquids.

Going further

Texts. Hansen & McDonald, Theory of Simple Liquids, 4th ed., Ch. 2 (distribution functions and the three routes). Pathria & Beale, Statistical Mechanics, 3rd ed., Ch. 10 (cluster expansions and correlation functions). Kardar, Statistical Physics of Particles, Ch. 5. Landau & Lifshitz, Statistical Physics Part 1, §116 (fluctuations and the compressibility sum rule). Barker & Henderson, Rev. Mod. Phys. 48, 587 (1976), for perturbation theory of liquids. The key numeric and symbolic results used throughout — the low-density limit $g\to e^{-\beta u}$ , the pressure-route reduction to $B_2=-\tfrac12\int f\,\mathrm{d}^3r$ by integration by parts, the ideal-gas compressibility $\kappa_T=1/(nk_BT)$ and $S(0)=1$ , the hard-sphere $B_2=\tfrac23\pi\sigma^3=4v_{\text{mol}}$ , and $B_2\approx-16$ cm³/mol for LJ argon at 300 K — are all consistent with one another. A note on the Carnahan–Starling equation: the correct consistency interpolation is $\tfrac13 P_{\text{vir}}+\tfrac23 P_{\text{comp}}$ (two-thirds compressibility, one-third virial), which reproduces $Z_{\text{CS}}=(1+\eta+\eta^2-\eta^3)/(1-\eta)^3$ ; the reversed weighting $\tfrac13 P_{\text{comp}}+\tfrac23 P_{\text{vir}}$ does not.

Prerequisites. Configuration integral · The Mayer f-function · Virial expansion of the equation of state.

Leads to. Second virial coefficient · Maxwell construction and coexistence · Irreducible cluster integrals and the virial series.

Why does the compressibility equation contain no pair potential, while the energy and pressure equations do?

Because it comes purely from number-fluctuation statistics in the grand canonical ensemble. The variance $\langle(\Delta N)^2\rangle$ is fixed by the two-particle density $\rho^{(2)}=n^2 g(r)$ — that is, by how particles are distributed — regardless of what forces produced that distribution. The energy and pressure equations, by contrast, average $u(r)$ and $r\,u'(r)$ over pairs, so they explicitly assume the total energy is a sum of pair terms. This is why the compressibility route remains exact even for genuinely many-body potentials, whereas the other two routes do not.

Why do the three routes disagree when I use an approximate g(r)?

Because the exact $g(r)$ satisfies an entire infinite hierarchy of relations (the BBGKY hierarchy plus closure conditions), and those relations are exactly what force the three routes to give one consistent equation of state. An approximate closure such as Percus–Yevick or the hypernetted chain satisfies only part of that hierarchy, so each route emphasizes different features of the (imperfect) $g(r)$ and they no longer agree. The gap between, say, the virial and compressibility pressures is a direct, quantitative measure of how far the approximate $g(r)$ is from the truth — and schemes like the modified HNC tune a free parameter precisely to close that gap.

How is g(r) actually measured?

Through scattering. A beam of X-rays or neutrons fired at a liquid scatters with an intensity proportional to the static structure factor $S(k)$ , which is directly related to $g(r)$ by the Fourier transform $S(k)=1+n\int[g(r)-1]e^{-i\mathbf k\cdot\mathbf r}\,\mathrm{d}^3r$ . Measuring $S(k)$ over a range of scattering angles and inverting the transform yields the experimental $g(r)$ . The first peak tells you the nearest-neighbour spacing, its area the coordination number, and the tiny-angle ( $k\to0$ ) intensity the isothermal compressibility. For colloids, where particles are large enough to image, one can even build $g(r)$ by directly histogramming particle separations from microscope snapshots.

What is the difference between g(r) and the total correlation function h(r)?

They differ by a constant: $h(r)=g(r)-1$ . The point of subtracting the 1 is that $g(r)\to1$ at large $r$ (uncorrelated), so $g(r)$ itself does not decay, but $h(r)\to0$ — it is the genuinely short-ranged (away from criticality) excess correlation. The compressibility integral is naturally written with $h(r)$ : $nk_BT\kappa_T=1+n\int h(r)\,\mathrm{d}^3r$ . The Ornstein–Zernike relation goes one step further, splitting $h(r)$ into a "direct" correlation $c(r)$ plus indirect chains, which is the starting point for the integral-equation closures.

Why is the low-density limit of g(r) the Boltzmann factor e^{−βu}, not just 1?

At vanishing density, a given pair of particles is essentially alone — no third particle is close enough to matter. For an isolated pair the relative probability of a separation $r$ is set by the Boltzmann weight of their interaction energy, $e^{-\beta u(r)}$ . So $g(r)\to e^{-\beta u(r)}$ : it dips to zero inside a hard core (infinite energy), rises above 1 in an attractive well (favourable energy), and returns to 1 where $u\to0$ . The correction $g(r)=e^{-\beta u}[1+n\int f f\,\mathrm{d}^3r_3+\cdots]$ brings in the influence of a third particle at order $n$ , which is what generates $B_3$ and the higher virial coefficients.

Does g(r) capture everything about a liquid's structure?

It captures everything a two-body potential needs, and everything that two-point scattering measures — but not more. $g(r)$ is a two-particle quantity, so it says nothing directly about three-body correlations (how a triangle of particles is arranged). For pairwise-additive potentials that is enough for the energy and pressure, because those are pair averages. But if the true interactions include genuine three-body forces (as in dense noble gases or many metals), the energy and pressure routes need the three-particle distribution too, and $g(r)$ alone is insufficient. The compressibility route still works, since it never needed the potential in the first place.