X-ray Crystallography

What X-ray crystallography does

Almost everything we know about the exact geometry of molecules — the 109.5° tetrahedral angle at a saturated carbon, the planar hexagon of benzene, the double-helix pitch of DNA, the fold of an enzyme wrapped around its substrate — was measured, at some point, by X-ray crystallography. It is the single most powerful method for determining where every atom sits in three dimensions, down to a few thousandths of an ångström.

The idea is a piece of physics turned into a ruler. You cannot photograph a molecule with visible light: a molecule is roughly 1–10 Å across, while visible light has a wavelength of ~5000 Å, thousands of times too coarse to resolve atoms. X-rays, however, have wavelengths of about 1 Å — the same scale as the bonds between atoms. That match is the whole trick. When you shine X-rays through a crystal, the regular lattice of atoms scatters the beam into a pattern of sharp spots, and that pattern is a coded image of the atomic arrangement. Decode it and you have the structure.

The subtlety is that a crystal does not form a real image the way a lens does — there is no X-ray lens. Instead you record the scattered diffraction pattern and reconstruct the image mathematically, by Fourier synthesis. That reconstruction is where the famous phase problem lives, and it is the reason crystallography was a Nobel-winning intellectual achievement rather than just a photograph.

The pipeline, step by step

A crystal structure determination is a fixed sequence of steps. Each one feeds the next:

1. Grow a single crystal. Coax the molecule out of solution into an ordered lattice — a few tens of micrometres is enough for a synchrotron. This is often the hardest and slowest step. No crystal, no experiment.
2. Illuminate and diffract. Mount the crystal, cool it (usually to 100 K in a nitrogen stream to suppress radiation damage and thermal motion), and rotate it in a monochromatic X-ray beam. Each lattice-plane family flashes a spot onto the detector whenever it satisfies Bragg's law.
3. Record intensities. As the crystal rotates through hundreds of orientations, a 2D detector captures thousands of reflections. Software indexes each spot to a triple of integers (h, k, l) — the Miller indices — and integrates its brightness to get the intensity I(hkl).
4. Reduce the data. Merge symmetry-equivalent reflections, correct for absorption and detector geometry, and convert intensities into structure-factor amplitudes |F(hkl)| = √I(hkl).
5. Solve the phase problem. Recover the missing phase angle φ(hkl) for each reflection — by direct methods, isomorphous replacement, anomalous scattering, or molecular replacement (see below). This is the crux.
6. Compute the electron density. With amplitudes and phases in hand, run the Fourier synthesis to build a 3D map ρ(x,y,z) of where the electrons are. Atoms show up as peaks in that map.
7. Build and refine the model. Drop atoms into the density peaks, then refine their positions and thermal parameters by least-squares (or maximum likelihood) until the calculated diffraction matches the observed data. The agreement is scored by the R-factor.

The physics that ties it all together is a Fourier transform. The electron density and the diffraction pattern are Fourier mates of each other:

    ρ(x,y,z) = (1/V) · Σ_hkl  |F(hkl)| · exp(iφ(hkl)) · exp(−2πi(hx + ky + lz))

      where  |F(hkl)|  = measured amplitude   (√ of the spot intensity)
             φ(hkl)    = phase                (NOT measured — the phase problem)
             V         = unit-cell volume
             (h,k,l)   = Miller indices of each reflection

Every reflection is one term in the sum — one 3D wave of a particular direction and spacing. Add all the waves together with the right amplitudes and phases and the atoms emerge from the interference. Get the phases wrong and the map is noise.

Bragg's law: where the spots land

The geometry half of the problem — where each spot appears — is governed by Bragg's law, derived by W. Lawrence Bragg in 1912 and stated as:

    n·λ = 2·d·sinθ

      λ = X-ray wavelength (e.g. Cu Kα = 1.5418 Å)
      d = spacing between a family of parallel lattice planes
      θ = the angle between the incoming beam and those planes
      n = an integer (the order of the reflection)

Picture two parallel lattice planes a distance d apart. An X-ray reflecting off the lower plane travels an extra path of 2·d·sinθ compared with one reflecting off the upper plane. When that extra path is a whole number of wavelengths, the two waves emerge in phase and interfere constructively — a bright spot. At any other angle they interfere destructively and you see nothing. So diffraction spots appear only at discrete Bragg angles, and each angle decodes a particular plane spacing d.

Because small d-spacings require large sinθ (larger scattering angles), high-resolution detail lives in the outer reflections. The smallest d you can still measure is the resolution of your structure. Bragg's law fixes only the spot positions — the size and symmetry of the unit cell. It says nothing about how bright each spot is; that brightness comes from the structure factor and holds the atomic coordinates.

The structure factor: where the atoms hide

Each reflection's amplitude and phase together make up its structure factor F(hkl), a sum over every atom j in the unit cell:

    F(hkl) = Σ_j  f_j · exp( 2πi (h·x_j + k·y_j + l·z_j) )

      f_j            = scattering factor of atom j (roughly ∝ its electron count Z)
      (x_j, y_j, z_j) = fractional coordinates of atom j in the cell

Two facts fall straight out of this equation. First, heavy atoms dominate the scattering: because f_j scales with the number of electrons, a single iodine (Z = 53) or a metal outscatters dozens of carbons. This is exactly why crystallographers deliberately soak in heavy atoms to attack the phase problem. Second, the intensity of a spot, I(hkl) ∝ |F(hkl)|², depends on where the atoms sit through those phase terms. Move an atom and you shift the interference, changing spot brightnesses across the whole pattern. Reading those brightnesses backwards is how you locate the atoms — provided you can also recover the phases.

The phase problem and how it is solved

Here is the central obstacle. The Fourier synthesis needs, for every reflection, both an amplitude |F| and a phase φ. The detector measures only intensity, I ∝ |F|². Taking the square root gives back |F|, but the phase φ is irrecoverably lost at the moment of measurement. Roughly half the information required to make the image is thrown away by the physics of detection. This is the phase problem, and defeating it is what most of a crystallographer's cleverness goes into.

• Direct methods (small molecules). For structures of up to a few hundred atoms, the phases are not truly free — electron density must be everywhere positive and concentrated into atom-shaped peaks. These physical constraints impose statistical relationships (Sayre's equation, the tangent formula) among the phases of strong reflections. Programs like SHELXD bootstrap a consistent phase set from the strongest spots. Herbert Hauptman and Jerome Karle shared the 1985 Nobel Prize in Chemistry for this.
• Isomorphous replacement (proteins). Soak the crystal with a heavy-atom compound (a mercury or platinum salt) that binds without distorting the lattice. The heavy atom perturbs the intensities in a way that, compared against the native crystal, pins down its own position — and from that reference the protein phases can be estimated. Max Perutz used multiple isomorphous replacement to crack haemoglobin.
• Anomalous dispersion (SAD/MAD). Tune the X-ray wavelength (at a synchrotron) to an absorption edge of an element such as selenium — often incorporated as selenomethionine — so its scattering acquires a small phase-shifted imaginary component. That anomalous signal breaks the phase ambiguity. MAD (multi-wavelength) and SAD (single-wavelength) anomalous diffraction are now the workhorses of protein crystallography.
• Molecular replacement. If a structurally similar model already exists (a homologous protein, or the same molecule in another crystal form), rotate and translate that model into the new unit cell until its calculated diffraction matches. Borrowed phases from the model start the map. This is the fastest route when a template is available and is how most new protein structures are solved today.

Instrument, sources, and real conditions

• X-ray source. Laboratory diffractometers use sealed-tube or rotating-anode sources: a copper target gives Cu Kα at λ = 1.5418 Å, molybdenum gives Mo Kα at 0.7107 Å (shorter wavelength, better for small dense inorganics). Synchrotrons deliver beams billions of times brighter and tunable in wavelength — essential for MAD phasing and for micro-crystals.
• Monochromatization. A graphite or multilayer monochromator, or Kβ filter, selects a single wavelength so Bragg's law has a defined λ.
• Detector. Modern instruments use photon-counting pixel-array detectors (Pilatus, Eiger) that read out fast and noiselessly, replacing the film and image plates of earlier decades.
• Cryocooling. Flash-cooling the crystal to ~100 K in a nitrogen cold stream slows radiation damage from the intense beam and sharpens the diffraction by reducing thermal atomic motion. A cryoprotectant prevents ice rings that would obscure spots.
• Crystal size. A good single-crystal edge of 50–200 µm suffices in the lab; synchrotrons work with crystals just a few µm across, and X-ray free-electron lasers push toward single sub-micron crystals via serial crystallography.
• Data completeness. You rotate through enough orientations to sample a full asymmetric wedge of reciprocal space — often collecting many symmetry-redundant measurements to average down noise, quantified by metrics like Rmerge and I/σ(I).

How it compares with other structure methods

	X-ray crystallography	Solution NMR	Cryo-EM	Neutron diffraction
Needs a crystal?	Yes — the bottleneck	No	No (vitreous ice)	Yes (usually large)
What it probes	Electron density	Nuclear spins in solution	Coulomb potential (electrons + nuclei)	Nuclear positions
Typical resolution	0.8–3 Å	Distance restraints, not a map	2–3 Å, now to ~1.2 Å	1–2 Å
Size limit	None (salt → ribosome)	≲ 40 kDa in practice	Best for large assemblies	Needs big crystals
Sees hydrogen?	Weakly (only at high res)	Yes (¹H is the primary nucleus)	Rarely	Yes — its specialty
Dynamics / flexibility	Static snapshot	Yes — timescales & motion	Multiple conformations resolvable	Static
Central difficulty	Crystallization + phase problem	Signal overlap at large size	Low contrast, sample prep	Weak flux, huge crystals

Worked example: reading NaCl and DNA

Table salt, the first structure ever solved. In 1913 W. Lawrence Bragg and his father W. Henry Bragg pointed X-rays at a sodium chloride crystal. From the positions of the reflections they extracted the cubic unit-cell edge, a = 5.64 Å, and from the pattern of which reflections were present and how bright they were, they deduced the arrangement: Na⁺ and Cl⁻ ions alternating on a face-centred cubic lattice, each ion octahedrally surrounded by six of the other. The startling conclusion — that solid NaCl contains no discrete "molecules" of NaCl, only an infinite ionic lattice — could not have come from chemistry alone. It won the Braggs the 1915 Nobel Prize in Physics; at 25, W. L. Bragg remains the youngest science laureate.

The double helix. In 1952 Rosalind Franklin and Raymond Gosling recorded "Photograph 51," the X-ray fibre-diffraction pattern of the B-form of DNA. Its bold X-shaped cross of reflections is the unmistakable signature of a helix; the spacing of the layer lines gave a helical repeat of 34 Å and a base-to-base rise of 3.4 Å — exactly ten bases per turn. Watson and Crick used those measurements, together with Chargaff's base ratios, to build the antiparallel double-helix model. It is the most consequential diffraction pattern in the history of biology.

The same pipeline scales all the way up. Max Perutz spent 22 years solving haemoglobin (68 kDa); today a well-behaved protein can be phased by molecular replacement and refined in an afternoon, and the Protein Data Bank holds over 200,000 experimentally determined structures — the overwhelming majority from X-ray crystallography.

Limitations and pitfalls

• You must get a crystal. Many molecules — membrane proteins above all — resist crystallization entirely, or grow only into crystals too disordered to diffract well. This, not the physics, is the rate-limiting step of the whole field.
• It's a static, averaged, lattice-bound picture. The structure is a time and space average over all the unit cells, frozen at cryogenic temperature. Flexible loops and disordered side chains simply vanish from the density, and crystal-packing contacts can nudge a conformation away from what the molecule adopts in solution.
• Hydrogens are nearly invisible. With only one electron, hydrogen scatters weakly and is usually placed by geometry rather than seen; locating protons precisely requires neutron diffraction.
• Twinning and pseudo-symmetry. Merohedrally twinned crystals overlay two orientations' diffraction, and false symmetry can mislead space-group assignment — classic ways to derive a wrong or over-idealized model.
• Radiation damage. The intense beam breaks bonds — disulfides and metal centres are especially vulnerable — progressively degrading the very structure you are measuring. Cryocooling and dose spreading mitigate but never eliminate it.
• Model bias. Molecular replacement can imprint features of the search model onto the map; over-refinement to a low R-factor without validation (Ramachandran outliers, clashscore) produces confident-looking nonsense. Always cross-check the omit map and the free R-factor.

Discovery and the people behind it

The field was born in a single 1912 experiment. Max von Laue, reasoning that a crystal's atomic spacing should act as a diffraction grating for X-rays, had Walter Friedrich and Paul Knipping shine an X-ray beam through a copper sulfate crystal — and out came a pattern of spots. This proved simultaneously that X-rays are waves and that crystals are periodic lattices of atoms. Von Laue received the 1914 Nobel Prize in Physics.

Within a year the Braggs, father and son, turned the discovery into a tool: W. L. Bragg formulated the simple reflection law nλ = 2d·sinθ, and together they built the first X-ray spectrometer and solved NaCl, diamond, and zinc blende. Their 1915 Nobel Prize launched structural science. Over the following century the method matured in the hands of pioneers such as Dorothy Hodgkin — who determined penicillin (1945), vitamin B₁₂ (1956, Nobel Prize in Chemistry 1964), and insulin (1969) — Max Perutz and John Kendrew (haemoglobin and myoglobin, Nobel 1962), and Hauptman and Karle, whose direct methods (Nobel 1985) automated small-molecule phasing. Each advance chipped away at the phase problem and expanded the size of what could be solved.

Where it matters

• Drug discovery. Structure-based drug design begins with a crystal structure of the target protein — often co-crystallized with a candidate inhibitor bound in its active site — so chemists can see exactly how to improve the fit. HIV protease inhibitors and countless kinase drugs were engineered this way.
• Materials and minerals. The atomic packing of alloys, ceramics, zeolites, battery cathodes, and superconductors is mapped by crystallography, connecting structure to conductivity, catalysis, and strength.
• Pharmaceutical quality control. Different crystal packings of the same drug molecule — polymorphs — can have different solubility and shelf life; crystallography identifies which polymorph is in the pill, a legally and clinically critical distinction.
• Absolute configuration. Using the anomalous scattering signal, crystallography can determine which enantiomer of a chiral molecule you have in hand — the definitive assignment of absolute stereochemistry.
• Fundamental chemistry. Bond lengths, angles, coordination geometries, and hydrogen-bonding networks archived in databases like the Cambridge Structural Database (over 1.2 million organic and metal-organic structures) underpin our entire quantitative picture of molecular shape.