A polished GaN wafer may contain threading dislocations that terminate at the surface, while a deformed metal may contain a dense near-surface network whose arrangement changes around a crack or contact. Ordinary secondary-electron imaging reveals topography but not necessarily those lattice defects. Electron channeling contrast imaging uses the orientation dependence of electron scattering in a crystal to make local lattice distortion visible in a scanning electron microscope. The image becomes crystallographic evidence only when the excited reflection, beam direction, deviation from Bragg condition, detector response, surface state, visibility depth, and competing topographic or compositional contrast are controlled.

ECCI images defect-perturbed electron channeling in a bulk or film surface region. When the incident SEM beam approaches a diffraction condition, the internal electron wave field redistributes relative to atomic planes and changes the probability that electrons scatter back out of the specimen. A dislocation strain field, stacking fault, twin, interface, bend, or low-angle boundary perturbs that channeling state and modulates the backscattered-electron signal. ECCI therefore differs from EBSD: EBSD records a diffraction pattern to determine phase and orientation, whereas ECCI scans a selected diffraction condition to form a real-space defect image. Electron channeling patterns help choose that condition but are not themselves ECCI micrographs.

The geometric entry point remains Bragg diffraction. For plane spacing $d_{hkl}$, electron wavelength $\lambda$, order $m$, and Bragg angle $\theta_B$,

but ECCI contrast requires a dynamical description of the wave field and backscatter generation rather than a ray satisfying this equation once. The excitation or deviation parameter $s_g$ describes departure of reflection $g$ from exact Bragg condition; its sign convention must be stated. Near a selected band edge, small lattice rotations and elastic displacement fields change $s_g$ locally, redistribute channeling, and create bright-dark contrast whose polarity and width vary as the beam is rocked through the condition.

Controlled ECCI turns orientation knowledge into a reproducible diffraction condition. In conventional channeling-contrast imaging, the operator may tilt and rotate until defects appear, but the active reflection and exact excitation are uncertain. Controlled ECCI, or cECCI, first measures the local orientation by EBSD or an electron channeling pattern, simulates accessible channeling bands, and moves the stage or beam to a chosen condition. An isolated strong reflection near two-beam excitation simplifies defect interpretation; intersections of several band edges create multi-beam contrast that can defeat invisibility analysis.

Stage coordinates are not crystallographic coordinates. Tilt, rotation, specimen height, working distance, beam shift, scan rotation, detector placement, and mechanical backlash determine the realized incident direction. A nominal stage move calculated from one map point may miss the condition elsewhere because of specimen curvature, mounting error, grain rotation, or scan distortion. Fiducials and a post-move EBSD pattern or local rocking response should verify the actual orientation.

Rocking the incident direction across the selected band records the background BSE response and the evolution of defect contrast. Exact $s_g=0$ is not always the image with the thinnest or strongest feature; a small positive or negative deviation may be preferable depending on convention, detector, material, and geometry. The chosen working point should be recorded relative to the measured rocking curve rather than described only as “near Bragg.” Repeating $+g$, $-g$, and additional nonparallel reflections tests contrast polarity and defect visibility.

Define the defect type, density, crystallography, depth, and area needed
  -> Choose ECCI, EBSD, CL, EBIC, TEM, or a correlative sequence
  -> Prepare a flat low-damage surface and map topography, oxide, and contamination
  -> Acquire orientation and phase by EBSD or a calibrated channeling pattern
  -> Simulate accessible reflections and stage or beam moves
  -> Select an isolated strong g near a two-beam condition
  -> Set beam energy, convergence, current, working distance, detector, and dwell
  -> Verify the local rocking response and record the chosen deviation parameter
  -> Acquire ECCI with raw detector channels and contextual SEM images
  -> Repeat +g, -g, and independent reflections for visibility analysis
  -> Separate defect contrast from topography, composition, charging, and drift
  -> Quantify resolution, visibility depth, field selection, overlap, and count uncertainty
  -> Register EBSD, TEM, CL, EBIC, chemistry, or mechanical measurements
  -> Test the structure-property hypothesis with matched process controls
  -> Archive orientation, geometry, patterns, images, corrections, and provenance

Surface preparation and beam-detector settings determine whether channeling survives. ECCI needs a surface that is flat over the field and crystalline close enough to the surface for the incident wave field and escaping BSEs to retain diffraction sensitivity. Mechanical deformation, polishing relief, oxide, contamination, redeposition, FIB damage, charging, implanted ions, and residual stress can suppress or imitate defect contrast. Colloidal, electropolished, broad-ion-polished, cleaved, or epitaxial as-grown surfaces each have different benefits and artifacts. A final preparation step should be qualified on a known defect or channeling response, not only visual smoothness.

ECCI is often non-sectioning and far less preparation-intensive than TEM, but “non-destructive” needs qualification. Electron dose can charge, contaminate, heat, or damage sensitive materials. Ion polishing and FIB expose or modify a layer. Repeated tilting can change mechanical or environmental conditions in an in-situ experiment. Pre/post images, low-dose trials, and stability checks are necessary for semiconductors, battery compounds, oxides, two-dimensional materials, and reactive surfaces.

Beam energy changes penetration, scattering, channeling strength, probe size, defect visibility depth, and surface sensitivity. Higher current improves signal but may increase convergence or dose. A small convergence angle preserves a narrow incident-direction distribution; a high-resolution probe preserves lateral detail. Working distance and aperture must balance both. There is no universal best voltage or current across high- and low-atomic-number materials, film thicknesses, defect depths, and detector geometries.

A solid-angle-efficient BSE detector with useful angular selectivity is central. Detector position can emphasize true backscattered or forescattered components, and segmented detectors allow difference or sum images that suppress some topography. Gain, offset, saturation, dwell, scan speed, line averaging, frame registration, and post-processing determine apparent contrast. Raw segment signals should be retained because an aesthetically optimized combined image may conceal which angular electrons created the feature.

Dislocation identification requires multiple reflections and qualified invisibility criteria. Under an idealized two-beam condition, dislocation contrast depends strongly on diffraction vector $g$, Burgers vector $b$, displacement field, line direction $u$, depth, deviation parameter, and elastic anisotropy. A commonly used first invisibility condition is

with an additional condition involving $g\cdot(b\times u)$ for full invisibility in appropriate diffraction-contrast treatments. These are model-based criteria, not automatic truth tests. Other excited reflections, surface relaxation, mixed character, inclined lines, overlapping strain fields, curvature, and signal integration can leave residual contrast when a simple criterion predicts invisibility.

One disappearing feature under one $g$ does not uniquely determine $b$. A robust analysis collects several independent reflections, records $s_g$ and contrast polarity, lists crystallographically allowed Burgers vectors, and compares all visibility outcomes. $+g$ and $-g$ images can help interpret contrast reversal and dipoles. Image simulations or a TEM check strengthen assignments in complex materials. Failed or ambiguous conditions are evidence and should not be omitted.

Stacking faults and twins can form extended contrast because the lattice translation or orientation changes across a plane. Their fringe spacing, polarity, termination, and response to $g$ can distinguish them from slip traces or surface steps, but topography and composition must be checked. Low-angle boundaries may appear as arrays of dislocations or broad orientation contrast. Grain boundaries and phase interfaces can overwhelm nearby weak defect contrast because the active channeling condition changes across them.

Contrast and defect density are conditional observables rather than direct totals. A normalized image contrast can be defined as

for intensities extracted with a declared background and detector combination. Contrast depends on diffraction condition, filtering, gain, local orientation, surface slope, dwell, and depth, so values from different fields are comparable only after normalization and controls. Resolution should be separated into probe resolution, apparent feature width, minimum separable defects, and localization accuracy. A broad strain-field feature is not the dislocation-core diameter.

Visibility depth is finite and material dependent. Defects too deep contribute weak or broadened contrast; near-surface relaxation changes their strain fields; film interfaces truncate the channeling volume. Monte Carlo transport alone does not capture dynamical diffraction, while a two-beam model may simplify a multi-beam experiment. Voltage series, controlled-depth defects, cross-sectional TEM, or simulation can bound the effective depth. Quoting one depth from another material and voltage is unsafe.

For threading defects intersecting a surface, a visible areal count may be written

but the total density requires corrections for invisibility under the selected $g$, missed weak features, overlapping defects, field-edge rules, detector threshold, sampled depth, and field-selection bias. For line networks in a volume, projected line length per area is not automatically line length per volume. Multiple diffraction conditions reveal different subsets, so unioning registered detections and propagating duplicate-matching uncertainty is preferable to treating one image as complete.

Confidence intervals should respect spatial clustering. Dislocations arranged in cells, slip bands, low-angle boundaries, or bunches are not independent Poisson points. Multiple fields selected without looking at defect contrast, spanning relevant wafer radii, dies, pattern densities, process splits, and lots, support population claims. A high-magnification showcase image demonstrates visibility but does not establish density or uniformity.

ECCI complements rather than replaces EBSD, TEM, and functional defect maps. ECCI images real-space defect contrast over bulk surfaces without thinning every field, offering larger contextual areas and easier repeated or in-situ observations than conventional TEM. TEM offers better spatial resolution, controlled transmitted diffraction, defect depth through tilting and stereoscopy, and atomic-scale imaging, but thin-foil preparation, relaxation, limited area, and free surfaces can change the structure. EBSD supplies orientation and lattice-rotation maps but generally does not render every individual dislocation directly.

HR-EBSD estimates relative elastic strain and lattice rotation from pattern shifts, while GND analysis estimates net curvature compatible with geometrically necessary dislocations. ECCI can show individual or grouped lines, including statistically stored structures invisible to curvature alone, but only when they satisfy the selected contrast condition and lie within the visibility depth. Comparing ECCI line arrangements with HR-EBSD rotation or stress fields tests whether an inferred continuum field is physically plausible.

Semiconductor defect analysis is strongest when crystallographic and electronic maps are registered. In GaN, AlGaN, SiC, GaP-on-Si, III-V heterostructures, oxides, and related epitaxial films, ECCI can map threading and misfit dislocations, stacking faults, domain boundaries, surface steps, and relaxation networks without preparing a TEM foil at every site. Orientation and reflection selection help distinguish Burgers-vector families, while film thickness and interface depth set which defects can be seen.

Cathodoluminescence and EBIC reveal whether a crystallographic defect changes radiative recombination or carrier collection; EDS or EELS constrains chemistry; AFM separates steps and pits; X-ray diffraction measures broader mosaicity and relaxation; and TEM verifies critical defect structure. Registration must include beam-interaction volumes and carrier diffusion, which can make a functional dark spot wider or displaced relative to the ECCI feature. Spatial coincidence alone does not establish causality without controls for surface, doping, strain, and neighboring defects.

A reproducible ECCI deliverable preserves surface preparation, phase and orientation, specimen axes, active $g$, stage and beam geometry, measured or simulated channeling pattern, deviation parameter, voltage, current, convergence, working distance, detector segments, dwell, raw frames, filtering, contrast definition, visibility criteria, depth evidence, counting rules, field selection, software, and correlative registration. It distinguishes a channeling image from an orientation map, a visible subset from total defect density, and a low-preparation method from a universally nondestructive one. Read ECCI through the diffraction-condition-surface-signal-visibility-depth-and-correlation lens.