Patterson Function: Deciphering Crystal Structures from Interatomic Vectors

In the quest to visualize the atomic machinery of life and matter, X-ray crystallography stands as a premier technique. It allows scientists to create detailed three-dimensional models of molecules, from complex proteins to novel materials. However, a fundamental obstacle, known as the 'phase problem,' stands between the raw experimental data and the final atomic map. The diffraction experiment captures the intensity of scattered X-rays but loses the crucial phase information required for direct image reconstruction. This article delves into one of the most ingenious solutions to this enduring challenge: the Patterson function. Developed by A. L. Patterson in 1934, this mathematical tool transforms diffraction intensities into a map not of atoms, but of the vectors separating them.

Across the following chapters, we will unravel the power of this remarkable concept. The "Principles and Mechanisms" chapter will explain how the Patterson function is derived as an autocorrelation of electron density, what its peaks represent, and how their properties reveal vital structural clues. Following this, the "Applications and Interdisciplinary Connections" chapter will demonstrate how crystallographers wield the Patterson function as a versatile tool to locate heavy atoms, orient molecular fragments, diagnose crystal imperfections, and even probe the nature of disordered materials. By the end, you will understand how this elegant map of interatomic vectors provides a crucial key to unlocking the secrets of molecular architecture.

Imagine you're an archaeologist who has found a beautiful, shattered piece of ancient pottery. You have all the pieces, but the original manual showing how they fit together is lost forever. This is the dilemma of the X-ray crystallographer. In a diffraction experiment, we measure the intensities of scattered X-rays, which are like the shattered pieces. What's lost is the crucial "phase" information—the instruction manual that tells us how to assemble these pieces back into a clear image of the molecule. This is famously known as the ​​phase problem​​.

So, what can we do? We are faced with a list of intensity values, $I(\mathbf{h})$, which are proportional to the squared amplitude of something called the ​​structure factor​​, $|F(\mathbf{h})|^2$. A direct path to the electron density map, $\rho(\mathbf{r})$, which would show us the atoms, requires the full, complex structure factor, $F(\mathbf{h})$, not just its squared magnitude. In 1934, the physicist A. L. Patterson had a brilliant and deceptively simple idea: "What if we just take the data we do have—these intensities—and perform the mathematical reconstruction (a Fourier transform) anyway?" What picture would emerge from this act of beautiful desperation? The result is the ​​Patterson function​​, and the map it generates is a remarkable object that, while not a direct picture of the molecule, holds the keys to unlocking its structure.

The Patterson function, $P(\mathbf{u})$, is formally the Fourier transform of the intensities. But what does that mean? Thanks to a beautiful piece of mathematics called the convolution theorem, we know that the Patterson function is equivalent to the ​​autocorrelation of the electron density​​.

Let's unpack that. Imagine you have a transparent map of the electron density of a single unit cell of the crystal. Now imagine you make an identical copy of that map. The autocorrelation, $P(\mathbf{u})$, is a measure of how much the two maps overlap when you shift one copy relative to the other by a vector $\mathbf{u}$.

When the shift vector $\mathbf{u}$ is zero, the two maps are perfectly aligned. Every atom's electron cloud sits right on top of its twin, giving a perfect overlap. This results in an enormous peak at the map's origin ($\mathbf{u}=\mathbf{0}$). This ​​origin peak​​ is a universal feature of every Patterson map. It is the sum of every atom's correlation with itself—a collection of all possible zero-length vectors. While it's always the largest peak, it doesn't tell us anything about how different atoms are arranged relative to each other, so we usually ignore it.

The real treasures are the ​​non-origin peaks​​. Suppose there is a peak at some vector $\mathbf{u}_1$. This means that if you shift the copy of your electron density map by the vector $\mathbf{u}_1$, you get a significant overlap. When would this happen? It would happen if an atom in the original map lands on top of a different atom in the shifted map. In other words, a peak at $\mathbf{u}_1$ tells you that there is a pair of atoms in your crystal separated by exactly that vector, $\mathbf{u}_1 = \mathbf{r}_j - \mathbf{r}_i$.

So, the Patterson function is not a map of atomic positions. It is a ​​map of all interatomic vectors​​ within the crystal's unit cell. For a molecule with $N$ atoms, the Patterson map will contain peaks for all $N \times N = N^2$ possible vectors connecting any atom $i$ to any atom $j$. It's a comprehensive, and often very crowded, catalog of every distance and direction between every pair of atoms in the structure. One other elegant property is that for every vector $\mathbf{u}$, there's also a vector $-\mathbf{u}$ (from atom $j$ to atom $i$ instead of $i$ to $j$). This means the Patterson map is always ​​centrosymmetric​​, it has a center of inversion at the origin, regardless of whether the crystal itself does.

Not all peaks in the Patterson map are created equal. The "height" or "weight" of a peak at vector $\mathbf{u} = \mathbf{r}_j - \mathbf{r}_i$ is proportional to the product of the scattering powers of the two atoms involved, which we can approximate as the product of their atomic numbers, $Z_i Z_j$.

Imagine a simple, hypothetical 1D crystal with two atoms in its unit cell, atom A at the origin ($x=0$) and atom B at position $x_B$.. The Patterson map will have:

1. An origin peak at $u=0$ from the A-to-A and B-to-B self-vectors. Its weight is proportional to $Z_A^2 + Z_B^2$.
2. A peak at $u=x_B$ from the A-to-B vector. Its weight is proportional to $Z_A Z_B$.
3. A peak at $u=-x_B$ (or $1-x_B$ in the unit cell) from the B-to-A vector. Its weight is also proportional to $Z_B Z_A$.

This simple relationship is incredibly powerful. A vector connecting two "heavy" atoms (like iron, $Z=26$, or gold, $Z=79$) will produce a much, much stronger Patterson peak than one connecting two light atoms (like carbon, $Z=6$, or oxygen, $Z=8$). This insight is the foundation of the ​​heavy-atom method​​, a classic technique for solving the phase problem. If you can soak a heavy atom into your protein crystal, its signature will be written in bright, bold letters all over the Patterson map, helping you to locate it.

Solving a Patterson map is like solving a complex puzzle. You have a list of vectors, and you need to deduce the original set of points that could generate them. This is not a trivial task. For a protein with thousands of atoms, the map can be an intimidating, overlapping mess of $N^2$ peaks.

Fortunately, nature sometimes gives us a cheat sheet: ​​crystal symmetry​​. If a crystal belongs to a certain ​​space group​​, its atoms are not arranged randomly; their positions are related by specific symmetry operations. This symmetry imprints a beautiful and fantastically useful pattern onto the Patterson map.

Let's take the example of the space group $P2_1$, common for protein crystals. One of its symmetry operations is a twofold screw axis. This rule dictates that for every atom at position $(x, y, z)$, there must be an identical, symmetry-related atom at $(-x, y + 1/2, -z)$.

What is the vector connecting this pair of symmetry-mates?

Look closely at that result! The middle component of the vector, $v$, is always $1/2$, no matter what the original coordinates $(x,y,z)$ of the atom were. This means that all the Patterson peaks corresponding to vectors between these symmetry-related pairs will not be scattered randomly throughout the map. Instead, they will all be concentrated on a single plane where the $v$ coordinate is $1/2$. This special plane is called a ​​Harker section​​. By looking only at this section, we can filter out a huge number of other peaks and focus on a much simpler pattern that directly tells us about the positions of the atoms in the crystal. Identifying these symmetry-constrained loci is a cornerstone of structure solution.

The Patterson method is not a panacea. If you have two different heavy atoms in your structure, the map becomes a confusing superposition of three different vector sets: peaks for atom 1 with itself, peaks for atom 2 with itself, and cross-peaks between atoms 1 and 2. Untangling this is a significantly harder puzzle.

An even greater challenge arises when all atoms in the structure have similar scattering power, for example in a nucleic acid or a pure organic molecule. In this case, the weights of all interatomic vector peaks ($Z_i Z_j$) are very similar, resulting in a "flat" map with poor contrast, making it nearly impossible to solve.

Here, a yet more subtle and beautiful physical phenomenon comes to our rescue: ​​anomalous" or "resonant" scattering​​. The scattering power of an atom is not actually a fixed constant; it depends slightly on the energy of the incoming X-rays. This dependence becomes dramatic when the X-ray energy is tuned to be very close to an "absorption edge" of a specific element—a characteristic energy at which that element loves to absorb X-rays.

Imagine your protein contains selenium atoms (which can be biosynthetically incorporated). Away from selenium's absorption edge, its scattering power is similar to that of, say, a phosphorus atom, and the Patterson map might be confusing. But if you tune your X-ray source to an energy right near the selenium edge, its scattering factor, $f_{Se}$, changes significantly, while those of all the other atoms (C, N, O, S, P) remain constant.

By collecting diffraction data at a few different energies around the edge and calculating ​​difference Patterson maps​​ (subtracting one map from another), we can create a map where almost everything cancels out except for the vectors involving the selenium atoms. The selenium-selenium and selenium-other-atom vectors suddenly light up, while the rest of the map goes dark. This technique, known as ​​Multiple-wavelength Anomalous Dispersion (MAD)​​, turns the Patterson function from a challenging puzzle into a precision tool, allowing us to pinpoint specific atoms and solve structures that were once intractable. It is a perfect example of how scientists, faced with a limitation, can dig deeper into the laws of physics to turn that limitation into a powerful new advantage.

So, we have this marvelous mathematical construction, the Patterson function, born from the squares of our diffraction amplitudes. We discovered that it’s not a direct map of the atomic city, but something far more curious: a map of all the whispers between atoms, a complete catalogue of the vectors that separate them. You might be tempted to think of this as a step backward—a convoluted puzzle instead of a clear picture. But as we shall see, this very puzzle holds the keys to deciphering matter's most intricate architectures. Its applications stretch from the very heart of biology to the frontiers of materials science, serving as a master key, a diagnostic tool, and a universal language of correlation.

The most celebrated role of the Patterson function is as a weapon against the great "phase problem" in crystallography. Imagine you are trying to locate a spy in a crowded city. You don't have their coordinates, but you have a special map that shows the precise vector from the spy to every known landmark. If you can find the landmarks, you can pinpoint the spy. This is the essence of the "heavy atom method." By introducing a single, very heavy atom (like mercury or uranium) into a protein crystal, we give ourselves a powerful reference point.

But the cleverest trick is that we don't even need other landmarks, because the crystal's own symmetry provides them! A crystal is a repeating pattern, and its symmetry operations create identical copies of our heavy atom at predictable locations. For instance, in the common space group $P2_1$, a screw axis symmetry relates an atom at $(x, y, z)$ to an identical one at $(-x, y + 1/2, -z)$. The vector connecting this pair of "twins" is therefore $(-2x, 1/2, -2z)$. Notice something remarkable? The $v$ coordinate of this vector is always $1/2$, no matter where the atom is! This means that all such special vectors lie on a single, specific plane in the 3D Patterson map. This plane is called a ​​Harker section​​. A crystallographer can simply slice open the Patterson map at $v=1/2$, find the prominent peak there, and from its coordinates $(u,w)$, immediately calculate the heavy atom's position $(x,z) = (-u/2, -w/2)$. The symmetry of real space creates a profound simplification in Patterson space, handing us the key to the entire structure.

This idea can be made even more powerful. In techniques like Multi-wavelength Anomalous Dispersion (MAD), we use X-rays tuned to an energy that makes specific atoms (say, selenium atoms substituting for sulfur) resonate. This makes their scattering "voice" different—it acquires a real part ($f'$) and an imaginary part ($f''$) that change with the X-ray wavelength. We can then craft a special kind of Patterson map that listens only to the chorus of these selenium atoms. The ​​Anomalous Difference Patterson​​, for example, is built from the differences in intensity between symmetry-related reflections, a signal that is dominated by the imaginary $f''$ component of the anomalous scatterers. Alternatively, a ​​Dispersive Difference Patterson​​ is built from intensity differences measured at two different wavelengths, which is primarily sensitive to the change in the real $f'$ component. Both maps filter out the noise from the thousands of other atoms, leaving a clean picture of the vectors between the selenium atoms themselves.

When several heavy atoms are present, the map becomes a superposition of all their interatomic vectors. But even here, symmetry can help us unscramble the message. In a centrosymmetric crystal, an atom at position $\mathbf{r}$ (relative to an inversion center at the origin) has a twin at $-\mathbf{r}$. The vector between this pair—a Harker vector—is simply $2\mathbf{r}$. Therefore, a strong non-origin peak at a position $\mathbf{u}$ in the Patterson map suggests the presence of a heavy atom at coordinates $\mathbf{r} = \mathbf{u}/2$. By identifying such peaks, a crystallographer can perform a beautiful piece of detective work, proposing heavy atom positions to bootstrap the phasing process.

What if we don't have just a single heavy atom, but a whole piece of the structure is already known? Suppose we are trying to solve the structure of a human enzyme, and we already know the structure of a similar enzyme from a mouse. We can use the mouse structure as a "search model" to solve the human one. This powerful technique is called Molecular Replacement. But how do you find where to place the search model in the new crystal's unit cell?

The problem has two parts: finding the model's orientation (rotation) and then its position (translation). The Patterson function is the perfect tool for the first step. Why? Because the Patterson map—a collection of all interatomic vectors—is brilliantly insensitive to translation. If you shift an entire object, the internal vectors within it remain unchanged. The first step, a ​​rotation function​​, seeks to find the correct orientation of the search model by rotating it in all possible directions and, for each orientation, checking how well its own calculated Patterson map overlaps with the experimentally measured one.

Imagine our known fragment is a small, rigid, square-shaped cofactor made of five atoms. Its internal vector set is fixed and unique. Our experimental Patterson map for the whole protein contains these vectors, but mixed in with millions of others and rotated by some unknown angle. The rotation search is like taking our little square of vectors and spinning it around, looking for the angle where its pattern clicks perfectly into place atop a matching pattern in the experimental map. The angle that gives the maximum overlap, or score, is the correct orientation of our cofactor in the new crystal. Only after this rotation is found do we proceed to the second step, finding its position. The Patterson function's invariance to translation allows us to neatly decouple the six-dimensional search problem (three rotation angles and three translation coordinates) into two simpler three-dimensional problems.

Beyond structure solution, the Patterson map is an indispensable diagnostic tool, like a doctor’s X-ray for crystals. It can reveal subtle and not-so-subtle problems that would otherwise confound the analysis.

One of the most dramatic signatures is that of ​​translational non-crystallographic symmetry (tNCS)​​. This occurs when the crystal's fundamental building block (the asymmetric unit) contains two or more identical copies of a molecule related by a simple shift, a vector $\mathbf{t}$. The total electron density is the sum of the density of the first molecule, $\rho_1(\mathbf{r})$, and the second, $\rho_1(\mathbf{r}-\mathbf{t})$. When we compute the autocorrelation of this sum, a remarkable pattern emerges. We get the self-Patterson vectors of each molecule, which form a cluster of peaks around the origin. But we also get two "cross-Patterson" terms: a complete copy of the self-Patterson cluster shifted to be centered at $+\mathbf{t}$, and another at $-\mathbf{t}$. The result in the Patterson map is striking and unmistakable: a single, exceptionally strong peak far from the origin, located at the precise coordinates of the translation vector $\mathbf{t}$. This peak represents the constructive addition of all the vectors connecting every atom in the first molecule to its identical partner in the second.

What’s more, this diagnostic sign comes with a beautiful, quantitative rule. If there are $M$ identical copies of the molecule related by tNCS, the height of any one of these tNCS cross-peaks will be exactly $1/M$ times the height of the gargantuan origin peak. A peak half the height of the origin means two copies; a third the height means three. It is a stunningly simple law emerging from a complex situation, allowing a crystallographer to immediately diagnose not only the presence of multiple copies but also how many there are.

The Patterson map also helps diagnose more subtle pathologies like ​​pseudosymmetry​​. This occurs when a crystal is, for example, truly monoclinic (with one unique axis) but its unit cell dimensions are very close to being orthorhombic (with three perpendicular axes), say, with a $\beta$ angle of $90.3^{\circ}$. The crystal is "pretending" to have higher symmetry than it actually possesses. The Patterson map sees both the reality and the illusion. It becomes a confusing superposition of the vector set from the true, lower symmetry structure and a "ghost" vector set arising from the approximate, higher symmetry. This generates a forest of ambiguous peaks of comparable height, making it exceedingly difficult to find the heavy atom substructure and solve the phases. Here, the Patterson map does not offer an easy solution but provides the crucial diagnosis of why the problem is so difficult, guiding the scientist toward more sophisticated strategies.

The power of the Patterson function is not confined to the highly ordered world of single crystals. Its fundamental nature as an autocorrelation function gives it a universal voice. Consider the exciting new class of 2D materials called MXenes. These are synthesized as stacks of ultrathin sheets, but the stacking is often imperfect. The layers are parallel, but randomly shifted and rotated with respect to one another—a condition known as ​​turbostratic disorder​​.

Trying to make a 3D Patterson map of such a material would be meaningless. However, we can think about the structure just along the stacking direction, $z$. The electron density along this axis is a series of spikes, one for each layer. But if the stacking is disordered, the spacing between adjacent layers is not constant. Instead, it's a random variable that fluctuates around some mean value $d_0$. The 1D Patterson function, $P(z)$, now describes the probability distribution of all possible inter-layer distances.

An X-ray diffraction experiment that looks only at the reflections along this stacking direction is effectively measuring the Fourier transform of this 1D Patterson function. As the elegant theory of paracrystals shows, the result is no longer a set of infinitely sharp Bragg peaks. Instead, the diffraction pattern consists of broadened peaks with a specific asymmetric shape. The exact shape of these peaks is a direct report on the statistics of the disorder. The asymmetry and width of the peaks tell us the variance, $\sigma_d^2$, in the interlayer spacing. This is a profound conceptual leap. We have gone from locating individual atoms in a perfect crystal to characterizing statistical disorder in a functional nanomaterial. It is the same fundamental physics, the same mathematics of autocorrelation, applied to a different kind of order.

From the precise clockwork of an enzyme to the statistical rustle of a messy stack of nanosheets, the Patterson function provides a window into the soul of a structure. It is a testament to the unifying power of physical principles, showing us how the simple act of correlating a structure with itself can reveal its deepest secrets, its hidden symmetries, and its informative imperfections.