Cathodoluminescence
When an electron beam deposits energy in a semiconductor, it creates excited carriers far above thermal equilibrium. Some carriers lose energy, diffuse, become trapped, or recombine nonradiatively; others recombine by emitting photons. Cathodoluminescence (CL) collects those photons inside an SEM, STEM, or dedicated electron-beam system and relates their wavelength, intensity, timing, and beam position to band-edge emission, alloy composition, strain, impurities, quantum confinement, and recombination-active defects.
CL is an electron-excited optical measurement whose contrast follows an entire carrier history. The beam establishes a three-dimensional generation distribution (G(\mathbf r;\mathbf r_b)), after which carriers thermalize, drift or diffuse, exchange with traps, and compete between radiative and nonradiative pathways. A simplified detected signal at beam position (\mathbf r_b) is
where ηₒₚₜ includes photon extraction, mirror collection, spectrometer transmission, and detector response. The light may be generated away from the primary energy-deposition volume because carriers move before recombination. A dark line can therefore indicate a nonradiative defect, carrier escape, absorption, shadowed collection, charging, or specimen geometry—not simply “fewer photons were generated by the beam.”
Spectral peak energy identifies a transition only after the optical axis is calibrated. Photon energy and wavelength are related by
A band-edge peak can shift with alloy composition, strain, temperature, carrier density, electric field, quantum confinement, and instrument calibration. Broad defect bands may contain several overlapping transitions; a Gaussian decomposition is not automatically a unique set of defects. Wavelength calibration, dark subtraction, detector linearity, grating order, slit width, spectral resolution, and the wavelength-dependent response of mirror, windows, grating, and detector determine whether spectra acquired on different days or instruments can be compared quantitatively.
Beam energy and material stack define a generation volume, not a single analysis depth. Higher accelerating voltage generally deposits energy deeper and across a larger lateral volume, with material-dependent backscattering and transmission. Lower voltage can emphasize near-surface layers but increases sensitivity to oxides, contamination, charging, and surface recombination. In a multilayer, generated carriers may cross an interface or be captured by a quantum well before emitting. Monte Carlo energy-deposition simulations and voltage-dependent spectra help test which layers contribute, but carrier thermalization and diffusion can broaden the true CL source beyond a deposited-energy map.
The number of generated electron–hole pairs per unit time is sometimes estimated from absorbed beam power:
where (I_b/q) is the incident electron rate, (E_0) is beam energy, (f_{\mathrm{abs}}) represents the absorbed fraction after backscatter and transmission losses, and εₑₕ is the mean energy required per generated pair. This is an excitation estimate, not a photon-yield equation. Only a fraction of generated carriers recombine through the measured radiative transition, and only a fraction of those photons reach the detector.
| CL acquisition or comparison | Information gained | Principal confounder | Semiconductor use |
|---|---|---|---|
| Panchromatic CL image | Fast map of total detected emission | Spectral mixing and detector response | Locate dark dislocations or bright inclusions |
| Monochromatic map | Spatial distribution of a chosen band | Band overlap and wavelength drift | Separate band-edge from defect emission |
| Hyperspectral CL | Full spectrum at every beam position | Dose, drift, data volume, and fit non-uniqueness | Map alloy or strain-related spectral shifts |
| Beam-voltage series | Changes excitation-depth weighting | Changing interaction volume and injection density | Distinguish surface and buried emission |
| Beam-current series | Tests linearity and state filling | Heating, screening, saturation, and damage | Separate intrinsic emission from injection effects |
| Temperature-dependent CL | Resolves thermal quenching and localization | Spectral drift, condensation, and stage stability | Study excitons, traps, and nonradiative activation |
| time-resolved CL | Measures recombination dynamics after pulsed excitation | Instrument-response convolution and carrier transport | Compare local lifetime pathways |
Radiative intensity is controlled by competing recombination rates. For free carriers in a simple direct-gap model, radiative recombination may scale as (R_{\mathrm{rad}}=Bnp). A trap-assisted Shockley–Read–Hall contribution is often expressed as
These relations explain why the same defect can change intensity nonlinearly with injection, doping, temperature, or surface condition. A lower CL signal may reflect stronger nonradiative recombination, but it can also arise from carrier escape, optical absorption, collection shadowing, or a transition shifted outside the detection band. Quantitative internal quantum efficiency requires more than raw counts: excitation, collection, spectral response, and competing pathways must be modeled or calibrated.
Spatial resolution combines probe size, energy deposition, carrier motion, and photon collection. A small SEM probe does not guarantee a comparably small luminescence source. Carriers can diffuse before recombination, and photons can be reabsorbed and re-emitted elsewhere. Conversely, quantum wells, surfaces, strong recombination centers, or carrier localization can confine emission. A useful schematic broadening relation is
but the terms need not be Gaussian or independent. Resolution should be demonstrated on a relevant boundary or structure under the reported beam energy, temperature, and injection—not inferred solely from the nominal probe diameter or pixel pitch.
Temperature changes the semiconductor and the measurement system simultaneously. Cooling can sharpen transitions, stabilize excitons, suppress phonon-assisted broadening, alter carrier diffusion, and deactivate or activate recombination channels. It can also cause specimen drift, charging, ice or hydrocarbon condensation, and changes in optical alignment. Thermal quenching is often modeled with activated competing rates, but a fitted activation energy is not automatically a unique defect level. Multiple pathways, carrier escape, phase changes, and temperature-dependent absorption should be tested before assigning a microscopic mechanism.
Electron dose can change precisely the defects and interfaces being studied. Irradiation may charge an oxide, screen internal fields, fill traps, create or anneal color centers, desorb species, deposit carbon, heat the interaction volume, or drive atomic displacement. A current series tests injection regime; repeated fast frames reveal temporal evolution; blanked-beam recovery tests reversibility. Hyperspectral maps are especially dose intensive because a spectrum is collected at every pixel. Drift-corrected frame summation is often safer than one long raster, provided the earliest and latest spectra are compared for change.
Time-resolved CL records emission after pulsed excitation and can separate fast and slow recombination components. A measured transient is the convolution of the material response with the electron-pulse width and detector timing response:
Multi-exponential parameters are descriptive unless tied to a kinetic model; carrier transport into and out of the observed region can mimic a recombination lifetime. Instrument-response measurement, repetition-rate checks, pile-up control, background, and global fitting across wavelength or temperature strengthen an interpretation.
Correlative measurements distinguish optical consequence from structural cause. CL can map radiative efficiency and transition energy; EBIC maps charge collection and nonradiative electrical activity; EDS or EELS constrains composition; EBSD, diffraction, Raman, or HRXRD constrains orientation and strain; TEM locates defects and interfaces. A CL-dark dislocation that is also EBIC-dark has a stronger recombination interpretation than a dark optical line alone. A spectral shift supported by composition and strain measurements is more credible than assigning every wavelength change to alloy fraction.
For semiconductor process learning, the central question is not “where is the sample bright?” It is “which radiative transition or recombination change remains after excitation volume, carrier transport, optical throughput, injection, temperature, charging, and dose are bounded?” Reading CL through that excitation-transport-recombination-and-calibration lens converts colorful emission maps into defensible evidence about semiconductor defects and electronic structure.