Partial Radial Distribution Function

In the vast landscape of materials, the states of matter that are crowded but not perfectly orderly—the liquids and glasses—present a unique challenge. Unlike the predictable lattice of a crystal or the sparse randomness of a gas, their structure is a complex, dynamic dance of particles. How can we quantitatively describe and understand this "beautiful mess"? The key lies in a powerful statistical tool that acts as our microscope into the atomic world: the ​​partial radial distribution function (pRDF)​​. This article explores the pRDF as the fundamental language for describing the structure of disordered systems, addressing the gap between qualitative intuition and quantitative analysis of local atomic environments. Across the following chapters, you will gain a comprehensive understanding of this essential concept. The first section, ​​Principles and Mechanisms​​, will dissect what the pRDF is, how to interpret its features to reveal chemical ordering or clustering, and how it arises from the fundamental forces between atoms. Subsequently, the section on ​​Applications and Interdisciplinary Connections​​ will demonstrate how this structural map is used to calculate thermodynamic properties, predict material behavior, and even serve as a blueprint for designing new materials through computer simulation.

Imagine you could shrink yourself down to the size of an atom and sit on one, observing the world around you. What would you see? In a gas, other atoms would be few and far between, zipping by randomly. In a perfect crystal, you would see your neighbors arranged in a stunningly perfect, repeating lattice, stretching out to infinity in all directions. But what about the fascinating worlds in between—liquids and glasses? Here, things are crowded but not perfectly orderly. It is a bustling, dynamic, and intricate dance of particles. How can we make sense of this beautiful mess?

The ​​partial radial distribution function​​, or ​​pRDF​​, is our mathematical microscope for this world. It is a powerful tool that allows us to take a statistical snapshot of this atomic dance, providing a map of the average local environment around any given type of atom. This map, as we will see, is not just a jumble of numbers; it is rich with information, telling us stories of attraction, repulsion, order, and disorder.

Let's stick with our shrunken selves for a moment. Suppose we are in a liquid made of two types of atoms, say, A and B. We decide to sit on an A atom and look around. We pull out a special kind of radar that can only see B atoms. As we sweep this radar in all directions, we count how many B atoms we find in a thin spherical shell at a distance $r$ from us. We do this over and over, averaging over many different A atoms and over time, to smooth out the instantaneous fluctuations.

The pRDF, denoted as $g_{AB}(r)$, is essentially the result of this measurement, but with a crucial normalization. It's the density of B atoms we find at distance $r$ divided by the average density of B atoms in the liquid as a whole. In essence, it answers the question: "At a distance $r$ from an A atom, is it more or less likely to find a B atom than pure random chance would suggest?"

The interpretation is beautifully simple:

• If $g_{AB}(r) = 1$, finding a B atom at this distance is exactly as likely as picking a random spot in the liquid. The presence of our central A atom has no influence at this distance.
• If $g_{AB}(r) > 1$, it is more likely to find a B atom here. This indicates some form of attraction or structural ordering that pulls B atoms into a "shell" around the A atom.
• If $g_{AB}(r) < 1$, it is less likely. This suggests repulsion or that other atoms are crowding out the B atoms at this particular distance.
• If $g_{AB}(r) = 0$, it is impossible to find a B atom there. This typically happens at very small $r$, because you can't have two atoms occupying the same space—their hard-core repulsion forbids it.

When we plot $g(r)$ versus distance $r$, we see a characteristic pattern. A series of peaks emerge, corresponding to the first, second, and third "coordination shells"—the nearest neighbors, the next-nearest neighbors, and so on. As we look farther and farther away ($r \to \infty$), the influence of our central atom fades completely, and the correlations vanish. At these large distances, the local density of B atoms just becomes the average bulk density, so $g_{AB}(r)$ must approach 1. This is a universal feature of any disordered, homogeneous system.

With a full set of maps for our binary A-B mixture—$g_{AA}(r)$, $g_{BB}(r)$, and $g_{AB}(r)$—we can become structural detectives, deducing the "social preferences" of the atoms.

Imagine an A-B alloy where there is a strong energetic preference for atoms to be surrounded by atoms of a different type. This is known as ​​chemical short-range ordering​​. What would our maps show? If we sit on an A atom, we would expect to be surrounded by a posse of B atoms. Consequently, the first peak in the $g_{AB}(r)$ map will be sharp and tall, signifying a high probability of finding B atoms as nearest neighbors. Conversely, finding another A atom nearby would be rare, so the first peak in $g_{AA}(r)$ would be suppressed, and by symmetry, the same for $g_{BB}(r)$. This local structure is the microscopic fingerprint of a compound-forming tendency in the material.

Now, consider the opposite scenario: ​​clustering​​. In this case, atoms prefer to be near their own kind. A atoms form "cliques" with other A atoms, and B atoms do the same. This behavior is at the heart of phase separation. If we were to measure the pRDFs for such a material, the story would be inverted. Around a central A atom, the nearest-neighbor shell would be rich in other A atoms, leading to a high first peak in $g_{AA}(r)$. Encounters with B atoms would be disfavored, so the first peak of $g_{AB}(r)$ would be small. By comparing the relative heights of the like-pair ($g_{AA}$) and unlike-pair ($g_{AB}$) peaks, we can directly distinguish between a material that wants to form an ordered compound and one that wants to separate into A-rich and B-rich regions.

To make this comparison quantitative, we need a baseline. What does a structure with no chemical preference look like? In such a ​​random mixture​​, the identity of a neighbor is determined purely by stoichiometry. If the mixture is 50% A and 50% B, then a central atom's neighbors should, on average, also be 50% A and 50% B. By comparing the measured coordination numbers—the number of neighbors of each type, found by integrating the area under the first peak of the pRDF—to the values expected for a random mixture, we can precisely quantify the degree of ordering or clustering.

We can seal this intuition with an elegant thought experiment. Imagine a hypothetical mixture where the A and B particles are physically identical in every way—same size, same mass, and critically, the same interaction forces between all pairs ($U_{AA} = U_{BB} = U_{AB}$). The only difference is a nominal "label." What must the pRDFs look like? From the perspective of any single particle, the forces it feels from its neighbors are independent of their labels. The structural environment it creates around itself cannot possibly depend on its own label, nor can it differentiate between its neighbors' labels. The physics is blind to the labels. Therefore, all three partial radial distribution functions must be absolutely identical: $g_{AA}(r) = g_{BB}(r) = g_{AB}(r)$. This limiting case confirms that any difference between the pRDFs is a direct consequence of differences in the underlying interatomic forces.

This brings us to the deepest question: why do these structures form? The pRDF is not arbitrary; it is a direct consequence of the fundamental forces between particles, described by their pair potential energy, $U_{ij}(r)$.

Let's consider the simplest possible case: a very, very dilute gas. The particles are so far apart that the chance of three of them being close enough to interact simultaneously is negligible. The only thing that matters is the direct interaction between any two particles that happen to approach each other. In this low-density limit, there's a beautifully simple relationship between the potential and the structure, given by the Boltzmann distribution. The probability of finding two particles at a separation $r$ is proportional to $\exp(-U_{ij}(r)/k_B T)$. This means the pRDF is simply:

where $k_B$ is the Boltzmann constant and $T$ is the temperature. Where the potential energy $U_{ij}(r)$ is low (an attractive well), the probability is high, and $g_{ij}(r)$ shows a peak. Where the energy is very high (a repulsive core), the probability is near zero, and $g_{ij}(r)$ vanishes.

In a dense liquid, life is more complicated. The force between two particles is not just their direct interaction; it's also influenced by the jostling and positioning of all the surrounding neighbors. We can think of this as an "effective" potential, which we call the ​​potential of mean force​​, $W_{ij}(r)$. This $W_{ij}(r)$ represents the free energy change in the system when two particles, $i$ and $j$, are brought from infinite separation to a distance $r$. It includes the direct $U_{ij}(r)$ potential plus the averaged, or "mean," effects of all other surrounding particles. The wonderful thing is that the elegant relationship from statistical mechanics holds:

The pRDF is therefore a direct window into the free energy landscape of particle interactions in a dense system. The peaks in $g(r)$ correspond to the valleys in the potential of mean force—the stable, low-energy configurations for particle pairs.

The power of the pRDF extends beyond liquids into the realm of ​​amorphous solids​​, or glasses. These materials are solid-like but lack the periodic lattice of a crystal. They are, in essence, liquids frozen in time. The pRDF is the primary language we use to describe their structure.

In this context, we can distinguish between two fundamental types of disorder:

• ​​Topological Disorder​​: This is the inherent geometrical "messiness" of a non-crystalline network. In a perfect crystal, all bond lengths and angles are fixed. In a glass, these values are distributed around an average. A bond that is supposed to be 2.5 Å might be 2.4 Å here and 2.6 Å there. This distribution of bond lengths and angles is what causes the peaks in the pRDF to be broad, rather than infinitely sharp delta-functions. The more topological disorder, the broader the peaks. This is the fundamental signature of the amorphous state itself.

• ​​Chemical Disorder​​: This refers to having the "wrong" type of neighbor, relative to the material's preferred chemical bonding. Imagine a glass like silica ($\text{SiO}_2$), which strongly prefers to form Si-O-Si linkages. A state of perfect chemical order would have zero Si-Si or O-O bonds. If we find such "homopolar" bonds, we have chemical disorder. How would this appear in the pRDFs? It wouldn't necessarily broaden the peaks, but it would change their relative importance. The appearance or growth of a first peak in $g_{SiSi}(r)$ would be a direct signature of chemical disorder, indicating that some Si atoms have other Si atoms as nearest neighbors. This must come at the expense of Si-O bonds, so the coordination number under the $g_{SiO}(r)$ peak would decrease.

By carefully analyzing both the width (topology) and the area (chemistry) of the peaks in the various pRDFs, materials scientists can build a remarkably detailed picture of the structure of these complex materials, which is essential for designing everything from stronger metallic glasses to more transparent optical fibers.

In our previous discussion, we became acquainted with the partial radial distribution function, $g_{\alpha\beta}(r)$, as a precise mathematical tool for describing the structure of liquids and other disordered systems. We saw how its peaks and valleys paint a detailed picture of the atomic landscape, revealing where particles are likely to be found relative to one another. But knowing the map of a city is one thing; using that map to navigate, to understand its traffic flow, its social dynamics, and its economy is another entirely. Now, we shall embark on a journey to see what this remarkable function is good for. We will discover that the pRDF is far more than a static snapshot; it is a dynamic key that unlocks a deep understanding of a material’s properties and behavior, bridging the microscopic world of atoms with the macroscopic world we experience.

The most direct and intuitive application of the pRDF is to simply take a census of a particle's local environment. If you imagine yourself as an atom in a liquid, the pRDF tells you the average number of neighbors of a certain type you can expect to find at any given distance. By integrating the pRDF over the extent of its first major peak—what we call the first coordination or solvation shell—we can calculate the ​​coordination number​​. This tells us, on average, how many nearest neighbors a particle has.

For a simple liquid, this might be a single number. But in a mixture, say of atoms A and B, we can ask more specific questions. How many A's surround a central A? How many B's? The pRDF lets us answer each of these questions precisely by integrating the corresponding function—$g_{AA}(r)$ or $g_{AB}(r)$—over the first shell. For more complex systems, like a ternary mixture, this principle extends naturally, allowing us to meticulously catalogue the local environment of any given species.

This goes beyond mere counting. It allows us to characterize the local composition, which can be dramatically different from the overall bulk composition. Think of dissolving salt in water. The bulk mole fractions of ions might be small, but in the immediate vicinity of a positive sodium ion, there will be a strong preference for the negative ends of water molecules or for negative chloride ions. Using the coordination numbers for each species in its vicinity, we can calculate a local mole fraction. This phenomenon, known as preferential solvation, is fundamental to chemistry. The pRDF gives us a quantitative measure of this effect, explaining why chemical reactions can run at different rates in solution than one might expect from bulk concentrations alone. The local environment is what truly matters, and the pRDF is our guide to it.

A wonderful thing about physics is that seemingly distinct concepts are often intimately related. If we know the forces between particles—the potential energy function $u(r)$—and we know their average arrangement—the radial distribution function $g(r)$—we should be able to calculate the total energy of the system.

Indeed, this is the case. The average potential energy of a single particle is found by summing up the contributions from all its neighbors. The pRDF tells us exactly how many neighbors are at each distance $r$, so the total potential energy is simply an integral over all space of the pairwise potential energy $u_{\alpha\beta}(r)$ weighted by the number of pairs at that distance, which is given by the pRDF, $g_{\alpha\beta}(r)$.

This principle provides a powerful connection between microscopic structure and macroscopic thermodynamics. A beautiful example is found in the study of molten salts, like liquid sodium chloride. What holds such a liquid together? It is the electrostatic attraction between positive and negative ions. In the liquid, ions arrange themselves to maximize these attractions, a phenomenon called charge ordering, where each ion is preferentially surrounded by ions of the opposite charge. This ordering is perfectly captured by the difference between the unlike-ion and like-ion pRDFs, $g_{+-}(r) - g_{++}(r)$. By integrating this structural difference against the Coulomb potential, one can directly calculate the electrostatic cohesive energy of the entire liquid. The secrets of the liquid's stability are written in its structure, just waiting for the pRDF to reveal them.

At this point, you might be wondering if the pRDF is just a convenient theoretical fiction. How could we possibly measure the precise locations of atoms in a churning, chaotic liquid? The answer lies in the magic of scattering.

The primary way we determine the structure of materials—both crystalline and disordered—is by bombarding them with particles, typically X-ray photons or neutrons, and observing the pattern they make after scattering off the atoms. This is analogous to throwing a handful of pebbles into a pond and deducing the pattern of ripples on the surface.

In this world of scattering, the natural language is not that of real-space distance $r$, but of wavevectors $k$, which relate to the scattering angle. What an experiment measures is the ​​partial structure factor​​, $S_{\alpha\beta}(k)$. It turns out that the structure factor and the total correlation function, $h_{\alpha\beta}(r) = g_{\alpha\beta}(r) - 1$, are a Fourier transform pair. One can be calculated from the other. So, experimentalists measure $S_{\alpha\beta}(k)$ in "k-space" and then perform a mathematical transformation to obtain the $g_{\alpha\beta}(r)$ that we find so intuitive in "real space".

For a mixture with multiple components, a single experiment isn't enough, because the scattered signal is a weighted sum of all the partial structure factors ($S_{AA}$, $S_{BB}$, $S_{AB}$, etc.). The genius trick here is ​​contrast variation​​. By performing several different experiments where the "visibility" of each atomic species is changed, one can set up a system of linear equations and solve for each of the individual partial structure factors. This can be done by combining X-ray and neutron scattering (which see atoms differently) or, in the case of neutrons, by isotopic substitution (e.g., replacing hydrogen with deuterium), which changes an atom's scattering properties without altering its chemistry. In this way, we can experimentally eavesdrop on the atomic correlations and untangle the complete set of pRDFs for a complex mixture.

Armed with a tool that we can both understand theoretically and measure experimentally, we can start to use it for diagnosis. The pRDF serves as an exquisite fingerprint for the state of matter.

Let us consider the case of water. In its familiar crystalline form, ice Ih, every oxygen atom is locked into a periodic lattice, with precise, repeating distances to its neighbors near and far. The resulting oxygen-oxygen pRDF, $g_{OO}(r)$, reflects this perfectly: it displays a series of sharp, well-defined peaks that persist to very large distances, a clear signature of long-range order.

Now, compare this to Amorphous Solid Water (ASW), a glassy form made by freezing water vapor at very low temperatures. It lacks the periodic structure of a crystal. Its $g_{OO}(r)$ tells a different story. The first peak, corresponding to the nearest-neighbor hydrogen-bonded distance, is still there, but it is noticeably broader than in ice, reflecting the structural disorder. Beyond the second peak, the features are almost completely washed out, and the function decays to its baseline value of 1. The long-range order is gone, and the pRDF faithfully reports this fact.

What's more, the pRDF can distinguish between different types of structure. The intramolecular oxygen-hydrogen covalent bond within a water molecule is incredibly strong and stiff. If we look at the first peak of the $g_{OH}(r)$, which corresponds to this bond, we find it is extremely sharp and located at the same position for both ice and ASW. The molecule itself is unchanged; it is only their arrangement with respect to each other that differs. The pRDF is sensitive enough to make this crucial distinction between intramolecular and intermolecular structure.

The pRDF's power extends beyond static properties to the very dynamics of solution behavior, such as why some liquids mix and others, like oil and water, do not. The ​​Kirkwood-Buff theory​​ provides a breathtakingly elegant connection between the integrals of pRDFs and the thermodynamics of mixing.

The theory examines the "correlation volume" around a particle, which is the integral of the total correlation function, $g_{\alpha\beta}(r)-1$. This quantity, the Kirkwood-Buff Integral $G_{\alpha\beta}$, represents the net excess or deficit of particles of type $\beta$ in the vicinity of a particle of type $\alpha$, compared to a random distribution. By combining these integrals into a "thermodynamic factor," $\mathcal{M} = G_{AA} + G_{BB} - 2G_{AB}$, we can predict the tendency of the mixture. If particles have a strong preference for their own kind (large $G_{AA}$ and $G_{BB}$), $\mathcal{M}$ is positive, indicating a tendency to demix. If they prefer to associate with the other species (large $G_{AB}$), $\mathcal{M}$ is negative, indicating favorable mixing. The secret of miscibility is encoded in the integrated halos of correlation surrounding each particle.

This idea is further refined in the ​​Bhatia-Thornton formalism​​, which re-casts the scattering data into correlations of total number density and of concentration. As a mixture approaches a phase separation boundary, it develops large-scale fluctuations in its local concentration. These fluctuations are directly visible as a sharp, rising peak in the concentration-concentration structure factor, $S_{CC}(k)$, at small $k$ (long wavelengths). This macroscopic instability is a direct consequence of the underlying preferences encoded in the pRDFs.

Perhaps the most futuristic and powerful application of the pRDF lies not in analyzing existing materials, but in helping us design new ones. This is the world of computational modeling and the "inverse problem."

Suppose we want to create a computer simulation of a complex fluid, like a polymer solution or a biological membrane. To do this, we need to know the effective forces, or pair potentials $u(r)$, between the coarse-grained components. But these are often unknown. This is where the pRDF becomes a blueprint. We can begin by measuring the pRDFs of the real system using scattering experiments. This experimental $g(r)$ becomes our target structure.

Then, using the machinery of liquid-state theory, such as the Ornstein-Zernike equation combined with an approximation like the Hypernetted Chain (HNC) closure, we can work backward. These equations provide a link between the pair potentials $u(r)$ and the correlation functions $g(r)$. By starting with our experimental target $g(r)$, we can "invert" the equations to solve for the unknown potential $u(r)$ that must have been responsible for creating that structure.

This approach, known as structure-based coarse-graining, is a cornerstone of modern computational materials science. It allows scientists to develop simplified, yet physically accurate, simulation models directly from experimental data. It creates a beautiful feedback loop: experiment informs theory, which in turn builds better simulations, which can then predict the properties of new materials, guiding the next generation of experiments.

From a simple count of atomic neighbors to a blueprint for designing new materials, the partial radial distribution function has proven to be an indispensable concept. It is a unifying thread that weaves together thermodynamics, statistical mechanics, chemistry, and materials science, revealing the profound and elegant truth that in the world of matter, structure is the key to everything.