The dislocations are not wide, the imaging condition is (weak beam fixes both).
In a strong two-beam image a dislocation is a fat, fuzzy worm ~10 nm wide. Tilt the crystal so your reflection is weakly excited (the famous g–3g condition), and the same dislocation collapses into a crisp bright line ~1.5 nm wide on a nearly black background. This guide is built around the questions everyone actually asks about how and why that works, with live simulations to play with. Example material throughout: Cu.
Diffraction contrast has a natural length scale: the extinction distance ξg, the depth over which intensity sloshes between the direct and diffracted beams. At the exact Bragg condition that slosh is slow (ξ200 ≈ 27 nm in Cu at 200 kV), and a dislocation perturbs the image over ~ξg/3 ≈ 9–13 nm. Tilt away from Bragg and the slosh speeds up: the effective extinction distance shrinks as ξeff = ξg/√(1+(s·ξg)²). Sharper slosh → sharper image. Drag s and watch.
"g–3g" is a tilt recipe, read as: form the image with beam g, while tilting so that 3g sits exactly on the Bragg condition. That parks g far off Bragg (large sg), weakly excited, hence "weak beam", but in a reproducible, quotable way. Walk the steps below; the Ewald sphere, the systematic row, and the Kikuchi lines all move together.
This panel integrates the Howie–Whelan equations, the real dynamical two-beam theory, column by column through a foil containing one screw dislocation (line running into the screen). Far from the core the crystal is far off Bragg and stays dark; right beside the core the planes are locally bent back into the Bragg condition, and that narrow column lights up. That is the whole trick of weak beam. Play with s and watch the strong-beam worm collapse into a sharp line, and notice the bright line sits slightly beside the true core.
A dislocation only distorts planes that its displacement field actually bends: if the imaging reflection g is perpendicular to the Burgers vector b (so g·b = 0), those planes stay flat and the dislocation disappears. Flip through reflections and identify each segment's b: exactly the two-tilt experiment you'd run at the microscope. (Fine print: for edge dislocations a weak residual ghost survives unless g·(b×u) is also small.)
In fcc Cu every perfect dislocation is secretly two Shockley partials (b = ⅙⟨112⟩) bracketing a ribbon of stacking fault. Their separation d is set by a tug-of-war: elastic repulsion pushes them apart, the fault energy γ pulls them together, so measuring d measures γ. The catch: in Cu, d is only a few nm. A strong-beam image (width ~10 nm) shows one blurred worm; weak beam (width ~1.5 nm) resolves the pair. This is precisely the experiment Cockayne, Ray & Whelan invented the technique for in 1969.
Weak beam is dim by design: you deliberately threw away almost all the diffracted intensity to buy sharpness. Everything practical about the technique follows from that trade.
@misc{tripathy2026weakbeamdarkfield,
author = {Tripathy, Manisha},
title = {Weak-Beam Dark Field: Seeing Dislocations Sharp},
year = {2026},
howpublished = {\url{https://untetheredatom.com/tem/weak-beam-dark-field-guide}},
note = {Interactive teaching resource}
}