Radial distribution function

How do we describe the structure of a liquid, a glass, or any system that lacks the perfect, repeating order of a crystal? While a crystal's structure can be defined by a simple unit cell, the arrangement of atoms in a disordered material seems like a chaotic jumble. This article introduces the ​​radial distribution function (RDF)​​, a powerful mathematical tool that brings order to this chaos. It provides a universal language for quantifying the hidden architecture of matter by answering a simple question: for a given particle, what is the probability of finding another particle at a certain distance away? This article addresses the fundamental challenge of describing non-crystalline structures and reveals how the RDF serves as a bridge between the microscopic positions of atoms and the macroscopic properties we can observe.

Across the following chapters, we will build this concept from the ground up and explore its profound implications. The "Principles and Mechanisms" section will unpack the definition of the RDF, starting with the baseline of a perfectly random gas and introducing the structural features that arise from real interatomic forces. You will learn how to read an RDF plot and extract quantitative information, such as the number of nearest neighbors. Subsequently, the "Applications and Interdisciplinary Connections" chapter will showcase the remarkable versatility of the RDF, taking us from decoding the blueprint of glass and witnessing the quantum dance of electrons to connecting atomic arrangements with thermodynamic laws and even analyzing the structure of a living forest. We begin by exploring the fundamental rules that govern the atomic "social order" and how the radial distribution function captures them with elegant precision.

Imagine you're at a party. If it's a very sparse gathering in a huge hall, people might be scattered about with no rhyme or reason. The chance of finding someone standing three feet to your left is about the same as finding someone ten feet behind you. This is the picture of a completely random system, a state of perfect social anarchy. But what if the room is more crowded? People will naturally keep a certain "personal space." You won't find two people occupying the same spot. Friends will cluster together in groups. This subtle, unwritten set of social rules creates structure. The world of atoms and molecules behaves in much the same way, and the tool we use to map out these unwritten rules is the ​​radial distribution function​​.

To understand structure, we must first understand the complete lack of it. Let's consider the atomic equivalent of that sparsely populated party: a classical ​​ideal gas​​. In this theoretical gas, the "particles" are just points with no size and no forces between them; they are blissfully unaware of each other's existence. If we were to pick one particle and ask, "What is the probability of finding another particle at a distance $r$ away?", the answer would be completely independent of $r$. The local density of particles around our chosen one is simply... the average density of the whole system, $\rho_0$.

To quantify this, we define the ​​pair correlation function​​, $g(r)$, as the ratio of the local density at distance $r$ to the average density. For our ideal gas, since the local density is always the average density, we arrive at a beautifully simple result:

$g(r) = \frac{\rho_0}{\rho_0} = 1$

This is true for all distances $r$. This flat line at $g(r)=1$ is our North Star, our reference point for perfect randomness. It tells us there are no correlations; the position of one particle gives us zero information about the position of any other. Any deviation from $g(r)=1$ is a sign of structure, a whisper of the "social rules" governing the particles.

Now, let's introduce the most fundamental rule of the physical world: two things cannot be in the same place at the same time. Let's model our atoms not as points, but as tiny, impenetrable "hard spheres" of diameter $\sigma$. What does our function $g(r)$ look like now?

If you place one sphere at the origin, the center of another sphere simply cannot get closer than the distance $\sigma$. The probability of finding the center of another particle inside this "forbidden zone" is zero. Therefore, the local density is zero, and our pair correlation function immediately reflects this common-sense rule:

$g(r) = 0 \quad \text{for } r < \sigma$

This is the first feature we see in the structure of any real substance: a void for $r$ less than the particle size. This isn't just a mathematical quirk; it's the signature of the atom's own volume, its "personal space."

But what happens just outside this boundary, at $r = \sigma$? Because particles are crowded together in a liquid or solid, you are very likely to find a neighbor pressed right up against this boundary. This causes $g(r)$ to shoot up dramatically, creating a sharp, high peak. This first peak represents the first layer of neighbors, the "first coordination shell." As we move further away, we might see a second, smaller peak (the second layer of neighbors), and maybe a third, even fainter one. These peaks are like ripples in a pond, showing how the presence of one particle orders its neighbors, which in turn order their neighbors. Eventually, far away, the ripples die out, the influence of our original particle is lost, and $g(r)$ settles back down to 1. The system becomes random again. This oscillating decay to 1 is the classic fingerprint of the liquid state.

To work with these ideas, scientists use a few related functions. It's helpful to know the cast of characters:

• ​​The Pair Correlation Function, $g(r)$​​: As we've seen, this is a ​​dimensionless​​ quantity. It's a pure number that answers the question: "How much more (or less) likely am I to find a particle here compared to a perfectly random gas?" A value of $g(r)=2.5$ at the first peak means you are 2.5 times more likely to find a neighbor at that specific distance than you would by chance.

• ​​The Radial Distribution Function (RDF)​​: Often, what we really want to know is not just the relative probability, but the actual number of atoms at a certain distance. The volume of a thin spherical shell at radius $r$ with thickness $dr$ is $4\pi r^2 dr$. The number of atoms we expect to find in this shell is this volume multiplied by the local density, $\rho_0 g(r)$. We call this quantity the RDF, which we can write as $R(r) = 4\pi r^2 \rho_0 g(r)$. The units of $R(r)$ are atoms per unit length (e.g., atoms/Å). So, $R(r)dr$ is the literal number of particles in that thin shell.

• ​​The Reduced Pair Distribution Function, $G(r)$​​: Experimental techniques like X-ray scattering often yield a related function called $G(r)$, defined as $G(r) = 4\pi r \rho_0 (g(r) - 1)$. Since $g(r)$ oscillates around 1, $G(r)$ conveniently oscillates around 0, making the structural deviations from randomness immediately visible. It has units of inverse length squared (e.g., Å⁻²). Knowing any one of these functions allows you to calculate the others, for instance, $g(r)$ can be recovered from $G(r)$ using the simple rearrangement $g(r) = \frac{G(r)}{4\pi r \rho_0} + 1$.

This framework isn't just for making pretty graphs; it allows us to ask and answer one of the most fundamental questions about a material's structure: how many nearest neighbors does a typical atom have? This is the ​​coordination number​​.

Imagine the plot of the RDF, $R(r) = 4\pi r^2 \rho_0 g(r)$. The first big peak corresponds to that first shell of neighbors. To find the total number of atoms in this shell, we simply need to add up the atoms in all the thin concentric shells that make up that peak. In the language of calculus, we integrate. The first coordination number, $CN_1$, is the area under the first peak of the RDF, typically integrated from $r=0$ up to the first minimum ($r_{min}$) after the peak:

$CN_1 = \int_{0}^{r_{min}} 4\pi r^2 \rho_0 g(r) \, dr$

This is a profoundly powerful result. From a function derived from scattering experiments, we can compute a direct, intuitive, and integer-like quantity that describes the local bonding environment. For many simple liquids, this number is around 12, reminiscent of closely packed spheres. For a material like silica glass (SiO₂), we can use this method to confirm that each silicon atom is surrounded by about 4 oxygen atoms, revealing the fundamental tetrahedral building block of the glass.

There is an even deeper, more subtle rule hidden within the mathematics of $g(r)$. If we integrate the quantity $\rho_0 [g(r) - 1]$ over all space, we get a surprisingly simple answer: -1.

$\rho_0 \int [g(r) - 1] \, d\mathbf{r} = -1$

What does this mean? Think about it this way. The term $\int \rho_0 g(r) \, d\mathbf{r}$ counts the average number of other particles in the system, which is $N-1$. The term $\int \rho_0 (1) \, d\mathbf{r}$ represents the number of particles you'd find in the same volume if the system were completely random, which is $N$. The difference is $(N-1) - N = -1$.

This "sum rule" is a statement of particle conservation. It tells us that the "hole" created by the presence of the central particle (the region where $g(r) \lt 1$) is perfectly counterbalanced by the "excess" of particles in the surrounding shells (where $g(r) \gt 1$), with a net deficit of exactly one particle: the reference particle itself. It's a beautiful, self-consistent check on the entire framework.

The structure of a liquid is not a static photograph but a time-averaged snapshot of a frantic atomic dance. What happens if we increase the temperature? The atoms gain kinetic energy and jiggle around more violently. This thermal motion acts to smear out the delicate structure.

As a result, the sharp peaks in $g(r)$ will change shape. The first peak will become ​​lower​​ and ​​broader​​. "Lower" means the ordering is less pronounced; the probability of finding a neighbor at the most likely distance is reduced. "Broader" means the neighbors are found over a wider range of distances. The liquid's structure becomes more washed-out, more "gas-like." Conversely, cooling a liquid makes the peaks sharper and higher as the system settles into a more ordered arrangement, a precursor to freezing into a crystalline solid where the peaks would become infinitely sharp delta functions. Thus, $g(r)$ provides a direct window into the balance between the ordering effects of interatomic forces and the disordering effects of temperature.

So far, we have mostly imagined our particles as simple spheres. But the real world is filled with molecules of all shapes and sizes.

• ​​Complex Shapes​​: What about a liquid made of rod-like or disc-like molecules? Here, the interaction between two molecules depends not only on their distance but also on their relative orientation. A "collinear" arrangement might be very different from a "side-by-side" one. To capture this, the pair correlation function must be generalized to $g(\mathbf{r}, \mathbf{u}_1, \mathbf{u}_2)$, where $\mathbf{u}_1$ and $\mathbf{u}_2$ describe the orientations of the two molecules. This more complex function is essential for understanding the properties of liquid crystals, polymers, and even water, whose hydrogen bonds create strong orientational preferences.

• ​​Complex Compositions​​: What about materials made of more than one type of atom, like a salt solution or a metallic alloy? For a binary material made of atoms A and B, we can't just have one $g(r)$. We need to know the correlations for every possible pair. This gives rise to ​​partial pair distribution functions​​: $g_{AA}(r)$ (how A atoms arrange around other A atoms), $g_{BB}(r)$ (B around B), and $g_{AB}(r)$ (B around A). The total $g(r)$ measured in an experiment is a weighted sum of these partials. By cleverly using different types of radiation (like X-rays and neutrons) that scatter differently off A and B, scientists can disentangle these partials and build a complete, atom-by-atom 3D picture of even the most complex disordered materials.

From the simple baseline of an ideal gas to the intricate dance of atoms in a multicomponent glass, the radial distribution function provides a unified language for describing the hidden architecture of matter. It is a testament to how, by asking a simple question—"who are your neighbors?"—we can uncover the fundamental principles that govern the structure of almost everything around us.

Now that we have a feel for the principles of the radial distribution function, we can embark on a journey to see where this remarkable tool takes us. You might be surprised. We began by thinking about simple arrangements of particles, but we will soon find ourselves peering into the heart of glass, witnessing the strange quantum dance of electrons, and even analyzing the structure of a living forest. The function $g(r)$ is something like a universal Rosetta Stone; it allows us to translate the hidden patterns of nature, whether they are written in the language of physics, chemistry, or biology, into a simple, elegant graph that our minds can grasp. It reveals a beautiful unity in the way the world organizes itself.

Let's start with something you can hold in your hand: a piece of glass. It feels solid, yet physicists call it an "amorphous solid" or a "frozen liquid." Unlike a crystal, its atoms are not arranged in a neat, repeating lattice. It's a jumble. But is it a complete jumble? The radial distribution function is our key to finding out.

Imagine you are an atom inside a piece of glass. If you look at your immediate neighbors, you will find they are not just anywhere. In amorphous silica ($SiO_2$), for instance, a silicon atom will find itself surrounded by oxygen atoms at a very specific distance—the Si-O bond length. This is no accident; it is the dictate of chemistry. How do we know this for sure? We can't just look! But we can do a scattering experiment, bombarding the glass with X-rays or neutrons. The way they scatter gives us a pattern, encapsulated in a function called the structure factor, $S(Q)$. This pattern exists in a mathematical space that physicists call "reciprocal space," which is not very intuitive. The true magic happens when we perform a mathematical operation called a Fourier transform on this data. Doing so converts the scattering information from reciprocal space into the real-space information we crave, giving us the radial distribution function.

Suddenly, the jumble becomes clear. The $g(r)$ for amorphous silica shows a sharp, distinct first peak at about $1.61$ angstroms. This is the Si-O bond length, the fundamental building block of the glass. A second peak, around $2.62$ angstroms, tells us the distance between oxygen atoms within a single $SiO_4$ tetrahedron. A third, broader peak around $3.06$ angstroms reveals the distance between one silicon atom and the next, bridged by an oxygen. The $g(r)$ is like an architectural blueprint for the glass, revealing its short-range and medium-range order with stunning clarity.

The sensitivity of $g(r)$ is so great that it can distinguish between a true liquid and a glass. As a liquid is cooled so rapidly that it doesn't have time to crystallize, it becomes a glass. The atoms are locked into a disordered arrangement. This "locking in" of the structure manifests as a subtle but unmistakable feature in the radial distribution function: the second peak, which is a single broad hump in the liquid, often splits into two distinct sub-peaks in the glass. This splitting is a signature of more specific, geometric arrangements (like icosahedral clusters) that become frozen in place, a testament to the glass's frustrated attempt to find a more ordered state.

We can even push this technique further. In a material made of two or more types of atoms, like a metallic glass, we can ask: do atoms of type A prefer to be next to atoms of type B, or do they prefer their own kind? This "chemical short-range order" is crucial for determining a material's properties. By cleverly using different probes (like X-rays of different energies or neutrons), we can tease apart the total $g(r)$ into its constituent parts—the $g_{AA}(r)$, $g_{BB}(r)$, and $g_{AB}(r)$—and directly measure these chemical affinities. The RDF doesn't just show us where atoms are; it can tell us about their social lives!

Let's now shrink our perspective. We have been talking about atoms as if they were tiny billiard balls. But we know they are not. They are fuzzy quantum objects. What if we applied the idea of a radial distribution function not to a collection of atoms, but to the components within a single atom—specifically, to an electron?

The result is just as profound. In quantum mechanics, an electron in an atom doesn't have a fixed position; it exists in a cloud of probability. The radial distribution function for an atomic orbital, often written as $P(r)$, tells us the probability of finding the electron at a distance $r$ from the nucleus. For the 1s orbital of hydrogen, $P(r)$ starts at zero at the nucleus, rises to a single peak, and then trails off. That peak represents the most probable distance to find the electron—the Bohr radius! For the 2s orbital, something new appears: the RDF has two peaks, with a point of zero probability (a radial node) in between. The 3s orbital has three peaks and two nodes. The number of peaks in the electron's RDF is simply its principal quantum number, $n$. The abstract quantum numbers we learn in chemistry are made beautifully visible, representing shells of probability that an electron inhabits.

The quantum world, however, has an even stranger story to tell, and $g(r)$ is the perfect narrator. Let us consider not one electron, but two, in a sea of many. Is their arrangement correlated? You might think that, if we ignore their electric repulsion, their positions would be completely independent. But this is not true. The Pauli exclusion principle, a cornerstone of quantum theory, states that two identical fermions (like two electrons with the same spin) cannot occupy the same quantum state. If they are at the very same point in space, they must have different spins.

What does this mean for our pair correlation function? It means that for two electrons with parallel spins, the probability of finding them at the same location ($r=0$) is exactly zero. The function $g_{\uparrow\uparrow}(r)$ must start at $g_{\uparrow\uparrow}(0) = 0$. This isn't because of a repulsive force; it is a purely statistical effect born from the deep quantum nature of identity and indistinguishability. This region of vanishing probability around an electron where other, identical electrons are forbidden to enter is known as the ​​exchange hole​​ or ​​Fermi hole​​. It is as if every electron carries with it a private bubble of space, a halo of exclusion. In contrast, two electrons with opposite spins are not identical particles. The Pauli principle does not apply to them in the same way, and their pair correlation function, $g_{\uparrow\downarrow}(r)$, can be non-zero at $r=0$. In fact, for a non-interacting gas, $g_{\uparrow\downarrow}(r)=1$ everywhere. Using the tools of statistical mechanics, we can even calculate the precise mathematical shape of this hole, which turns out to depend on the density of the electrons in a beautifully intricate way. This is a direct, spatial picture of one of the most mysterious and powerful rules of quantum mechanics.

So, the RDF is a masterful descriptor of structure. But science is not just about description; it's about prediction. Can the microscopic arrangement revealed by $g(r)$ predict the macroscopic properties of a substance? The answer is a resounding yes.

Consider a simple glass of salt water. The positively charged sodium ions and negatively charged chloride ions are not distributed randomly. Each ion tends to be surrounded by ions of the opposite charge, forming a transient, flickering structure. The Debye-Hückel theory gives us an approximate mathematical form for the $g_{ij}(r)$ between these ions. Here comes the brilliant insight of Kirkwood-Buff theory: by performing a specific integral over this $g_{ij}(r)$, we can directly calculate macroscopic thermodynamic properties of the solution, such as its compressibility. The way tiny ions arrange themselves over distances of nanometers dictates how the entire solution responds when you squeeze it. The microscopic dictates the macroscopic.

This connection becomes even more dramatic when a system is on the verge of a profound change—a phase transition. Think of water at its boiling point, or a magnetic material at its critical temperature ($T_c$) where it loses its magnetism. At this critical point, the system is highly indecisive. Tiny fluctuations in one region can propagate and influence regions very far away. The correlations, which normally die off exponentially fast with distance, suddenly become long-ranged. How does $g(r)$ see this? At the critical point, the correlation function no longer decays exponentially. Instead, the correlation function decays according to a much slower power law. The correlation length, in theory, becomes infinite. The system acts as a single, coherent entity. The exponent describing this power-law decay is a "critical exponent," a universal number that tells us fundamental information about the nature of the transition, independent of the material's specific chemical details. The long-distance behavior of $g(r)$ becomes a window into one of the deepest and most beautiful areas of modern physics.

Our journey has taken us from the atomic to the quantum and back to the macroscopic. For our final stop, let us leap out of physics and chemistry entirely and land in the middle of a forest. Can the radial distribution function possibly have anything to say here?

Amazingly, it can. Think of the trees in a forest not as living things, but as points on a map. Are these points randomly located? Almost never. Some species may cluster together because of favorable soil, while others might be spaced far apart to minimize competition for sunlight. We can calculate a pair correlation function, $g(r)$, for the trees of a particular species. If $g(r) \gt 1$ for small $r$, it tells an ecologist that the trees are clustered. If $g(r) \lt 1$, they are over-dispersed.

But the connection goes deeper. This "microscopic" spatial pattern of individual trees has direct "macroscopic" consequences for the biodiversity of the entire forest. Imagine you are sampling the forest by throwing down small square plots and counting the number of species in each one (the alpha diversity). If a species is highly clustered, you are more likely to find many individuals in one plot and zero in another. This increases the chance that a randomly placed plot will miss that species entirely. Therefore, higher clustering (a larger peak in $g(r)$ at small $r$) leads to a lower average number of species found per plot. By using point process theory—the very same mathematical framework used in physics—ecologists can build models that predict how biodiversity metrics will change based on the spatial patterns of species, all quantified by their respective $g(r)$ functions.

From the structure of glass to the spatial statistics of a forest, the radial distribution function has proven to be a concept of astonishing power and versatility. It is a testament to the fact that deep, unifying principles underlie the workings of our universe, whether in the inanimate dance of atoms or the complex tapestry of life. It reminds us that by asking simple questions—"Where are your neighbors?"—we can begin to unravel the most intricate secrets of nature.