Powder vs Single-Crystal Diffraction

Same beam, same crystal — two ways to run the experiment

Guide 1 built the tool: a sealed tube pours out characteristic radiation, you filter it down to a single sharp line — copper K-alpha, wavelength lambda = 1.54 angstrom — and you have a clean monochromatic beam, an angstrom-scale ruler ready to measure atoms. The reciprocal-lattice rung told you what that ruler reads: a crystal scatters strongly only when the beam hits a plane at the special Bragg angle, and every such reflection is a point of the reciprocal lattice, the crystal's Fourier shadow. What guide 1 did not tell you is how to physically arrange the crystal and the detector so those reflections actually come out. That arrangement is what this guide is about.

Here is the geometric knot you must untie. A monochromatic beam fixes the radius of the Ewald sphere at 1/lambda, and a reflection only flares when a reciprocal point sits exactly on that sphere. But a stationary crystal in a fixed beam has almost no reciprocal points touching the sphere — a sphere is a thin surface, and the reciprocal lattice is a sparse grid of dots, so by sheer luck they rarely coincide. To collect reflections you must somehow bring point after point onto the sphere. There are two honest ways to do it, and they define the two great modes of X-ray diffraction.

Mode one: keep a single crystal but rotate it, so its reciprocal lattice turns and each dot sweeps through the sphere in its turn. Mode two: keep the beam and detector fixed, but replace the one crystal with a fine powder — a million tiny crystallites pointing every which way at once, so that whatever orientation the sphere needs, some grain is already offering it. The first is single-crystal diffraction; the second is powder diffraction. Same physics, same Bragg's law, wildly different data — and a real trade-off between them.

Single-crystal XRD: a full 3D map of every reflection

Picture a lone crystal — often smaller than a grain of salt, perhaps 0.1 to 0.3 mm — glued to a thin glass fibre and set spinning in the beam on a goniometer, a precision cradle that can tilt it to any angle. As the single crystal rotates, its whole reciprocal lattice rotates rigidly with it, and one after another the dots (hkl) pass through the Ewald sphere. Each crossing fires a sharp, separate spot onto the detector. Turn the crystal through a full set of angles and you harvest thousands of these spots, each one a distinct reflection labelled by its own (hkl).

The prize is that nothing is lost to averaging. Every reflection stands on its own, at its own place in three dimensions, carrying its own measured intensity — which, recall from the reciprocal-lattice rung, is the squared structure factor |F_hkl|^2, the motif's Fourier transform sampled at that dot. A single-crystal experiment therefore hands you the complete weighted reciprocal lattice: positions that pin down the unit cell, plus a full 3D list of intensities that encode where every atom sits. This is why single-crystal XRD is the gold standard for solving an unknown structure from scratch.

Powder XRD: a million crystals collapse 3D into 1D

Now grind the material to a fine powder — ideally crystallites a micron or so across, packed by the millions into a flat holder. Each grain is a perfect little crystal with its own reciprocal lattice, but the grains point in every direction at random. Think about what that does to a single reflection (hkl). In one grain its reciprocal dot sits up-and-to-the-left; in the next, off to the right; averaged over all orientations, that one dot is dragged all the way around the origin into a hollow sphere — a shell of radius 1/d_hkl. The crisp directional information is gone; only the distance from the origin, the magnitude 1/d_hkl, survives.

The Ewald sphere still has to be touched for diffraction to happen — but a shell and a sphere intersect in a whole circle, not a lone point. So instead of one spot, the diffracted rays for that (hkl) fan out into a hollow cone, its half-angle fixed by Bragg's law, and the detector catches it as a ring. Every d-spacing in the crystal throws its own concentric ring, and if you simply scan a detector outward and plot intensity against the diffraction angle 2-theta, all that 3D richness has collapsed into one tidy 1D curve — a diffractogram, a row of peaks marching across the 2-theta axis. This is the everyday output of powder XRD, the workhorse that runs on any polycrystal you can grind.

FROM ONE DOT TO A POWDER RING

 SINGLE CRYSTAL: reciprocal lattice = discrete dots
      . . . . .
      . . . . .    each (hkl) is ONE dot, at distance 1/d_hkl
      . . o . .     from the origin, in ONE direction
      . . . . .

 POWDER: average over ALL crystallite orientations
      the dot is dragged around the origin into a SHELL
            _.-""""-._
          .'          '.     shell radius = 1/d_hkl
         /   (origin)   \    (the DIRECTION is now lost;
         \      o       /     only the LENGTH survives)
          '._        _.'
             "-....-"

 Ewald sphere cuts each shell in a CIRCLE -> diffracted beam
 is a CONE (half-angle 2theta). Detector records rings, then
 a 1-D scan:

   I |     ||                    peak sits where
     |   | || |    |             sin(theta) = lambda / (2 d_hkl)
     |_|_|_||_|__|_|__ 2theta    POSITION <- d ,  HEIGHT <- |F|^2

Put a number on one peak. Take a plane with spacing d = 2 angstrom in a copper-source machine, lambda = 1.54 angstrom. Bragg's law gives sin(theta) = lambda / (2d) = 1.54 / (2 times 2) = 0.385, so theta = 22.6 degrees, and because a diffractometer reports the total deflection, the peak appears at 2-theta = 45.2 degrees. A smaller d — atoms more tightly stacked — pushes its peak to a larger angle; a larger d pulls it back toward the beam. That is the whole logic of a powder pattern's horizontal axis: every peak position is a d-spacing in disguise.

The trade-off: what powder averaging costs you

The two modes buy different things with different currency, and it is worth stating the exchange rate plainly. Powder is cheap and forgiving: no crystal to grow, just grind, pack, and press start — perfect for a metal, an alloy, a ceramic, a mineral, a pill. Single-crystal is dear and fussy: you must first coax nature into a good crystal. But look at what each keeps. Single-crystal preserves all three dimensions, so every reflection is resolved on its own. Powder throws two dimensions away, flattening the entire 3D map onto one axis.

That flattening has a sharp price: peak overlap. Two reflections with genuinely different (hkl) but nearly equal d-spacings land at nearly the same 2-theta and pile into one blurred peak — and in a high-symmetry crystal, several distinct planes can share exactly the same d and fuse into a single line. In single-crystal data those reflections would sit at separate points in space, cleanly apart; in a powder they are welded together, and you can no longer tell how much intensity belongs to each. That ambiguity is the fundamental reason powder data is harder to solve a structure from than single-crystal data — the same reason a shadow is harder to read than the object that casts it.

What each mode gives you, and where the ladder goes next

Even collapsed to 1D, a powder pattern is astonishingly informative, because its three visible features report on three different layers of the structure. The peak position is set by Bragg's law and so by the d-spacings — reverse that and you recover the unit cell, the step called indexing. The peak intensity is set by the structure factor and so by the motif — how heavy each atom is and where it sits inside the cell. And the peak width, the subtle one, is set by how big and how strained the crystallites are.

That last feature earns its own famous rule. Perfect, infinite crystals give razor-sharp peaks; as crystallites shrink toward the nanoscale, their peaks broaden, and the Scherrer equation t = K lambda / (beta cos theta) converts the extra width beta into a crystallite size t — a genuine ruler for nanocrystals read straight off a diffractogram. We will unpack all three readings — position, intensity, and width — properly in guide 3.

The most everyday use of all needs no structure solving whatever. Because every crystalline substance throws its own signature set of peak positions and heights, you can identify an unknown simply by matching its pattern against a database of known ones — phase identification, the fingerprinting that runs in labs from geology to pharma to failure analysis. It is fast, routine, and answers 'what is this?' without ever asking 'where is each atom?'.

But when you do want every atom, one deep obstacle remains no matter which mode you chose. A detector records only intensities, |F_hkl|^2 — the amplitude of each wave, never its phase. To Fourier-synthesise the electron-density map and actually see the atoms you need both, and half the information is thrown away the instant you measure. That missing half is the phase problem, the central puzzle of guide 4, tackled there by direct methods and the Patterson function. And guide 5 closes the rung by fitting a whole powder pattern at once with Rietveld refinement, squeezing a full structure out of even that flattened 1D curve. Powder or single-crystal, the road from here leads through the phase problem to a solved structure.