Photoluminescence spectroscopy asks what happens after a semiconductor absorbs light energetic enough to create excited carriers. Some electrons and holes recombine radiatively and emit photons whose energies, line shapes, and spatial distribution encode band structure, alloy composition, strain, temperature, and recombination pathways. The emitted spectrum is powerful precisely because it is indirect: the detector sees the combined result of carrier generation, transport, recombination, reabsorption, and the optical system.

The excitation and emission photons play different roles. The pump photon energy $E_{exc}=hc/\lambda_{exc}$ must be absorbed by an allowed transition or defect pathway, while the emitted photon reports a later recombination event after carriers have usually relaxed toward lower-energy states. For a band-edge feature,

but $E_{PL}$ is not automatically the unperturbed band gap. Exciton binding, alloy disorder, strain, quantum confinement, band filling, band-gap renormalization, temperature, and spectrometer calibration can shift the peak. Indirect-gap materials such as silicon additionally require phonon assistance for momentum conservation, so their spectra and efficiencies differ fundamentally from direct-gap III–V or many wide-bandgap emitters.

Steady-state intensity is governed by a generation–recombination balance. Under constant illumination, excess carrier density settles where optical generation equals all recombination channels,

The detected PL is proportional to the radiative recombination integrated over the excited and collected volume, multiplied by escape, collection, and instrument-response factors. A dark region can indicate stronger nonradiative recombination, but it can also result from lower absorption, shadowing, focus error, surface texture, reabsorption, or collection geometry. PL counts alone are therefore neither an absolute lifetime nor an absolute defect density.

Peak energy can measure alloy composition only through a validated calibration state. A composition-dependent band gap may be written in a bowing model such as

where $b$ is a material- and temperature-specific bowing parameter. Turning a fitted PL peak into mole fraction requires defined temperature, strain state, doping, excitation density, peak model, and reference materials. NIST studies of compound-semiconductor standards found that fitting method, measurement temperature, and doping concentration influence PL-based composition assessment. A quoted composition uncertainty must include those effects rather than only the wavelength repeatability.

Excitation density is a measurement axis, not a nuisance setting. Changing pump power changes carrier population and can saturate traps, alter surface recombination, fill localized states, heat the specimen, screen internal fields, or activate Auger loss. Power-dependent peak energy and integrated intensity help distinguish mechanisms; a local relation $I_{PL}\propto P^m$ is descriptive only over the reported range and geometry. The power at the specimen, spot profile, photon energy, chopping or duty cycle, dwell time, and absorptance should be recorded. Comparing materials at equal laser-dial percentage does not establish equal generation rate.

Temperature controls both the semiconductor and the spectrum. Band gaps normally move with temperature, carrier distributions broaden, traps change occupancy, and nonradiative rates can activate thermally. A frequently used empirical band-gap form is

where $\alpha$ and $\beta$ are fitted for a particular material and regime. Laser heating can make the illuminated volume warmer than the stage sensor. Power series, anti-Stokes or Raman thermometry where applicable, and stable cryostat or chuck measurements help distinguish specimen temperature from excitation-induced heating. Every reference and production wafer must be compared at a controlled, documented thermal state.

st=>start: Define measurand: band edge, composition, defects, relative quality, or uniformity
design=>operation: Select excitation energy, power range, spot size, temperature, and collection geometry
cal=>operation: Calibrate wavelength, spectral response, dark signal, linearity, and spatial response
ref=>operation: Measure reference specimen and excitation power at the sample
acq=>operation: Acquire background-corrected spectra across power and selected temperature
fit=>operation: Fit physically justified peaks with instrument broadening and residual checks
id=>condition: Are peak assignment and competing variables independently constrained?
aux=>operation: Add temperature, polarization, power dependence, absorption, Raman, or XRD data
map=>operation: Map fitted observables with focus, drift, revisit, and normalization controls
unc=>operation: Propagate calibration, fitting, excitation, optical-transfer, reference, and model uncertainty
out=>end: Report spectra, settings, observables, assumptions, spatial resolution, and uncertainty
st->design->cal->ref->acq->fit->id
id(yes)->map->unc->out
id(no)->aux->acq

Photoluminescence mapping must map fitted physics rather than raw brightness alone. A hyperspectral map can store peak energy, linewidth, band ratios, and integrated intensity at every position, while camera-based imaging trades spectral information for throughput. In either case, measured contrast is convolved with the excitation and collection point-spread functions. Step size finer than the optical resolution oversamples rather than creates new spatial detail. Wafer bow, patterned topography, illumination nonuniformity, vignetting, detector drift, and varying surface reflectance require focus control, flat-field or reference normalization, and repeated control sites.

External luminescence efficiency includes optical escape as well as internal recombination. Internal quantum efficiency compares photons generated inside the material with absorbed pump photons; external quantum efficiency compares photons leaving toward the measurement environment with incident or absorbed photons under a specified definition. Reflection, parasitic absorption, total internal reflection, reabsorption, photon recycling, and collection solid angle separate the two. Absolute measurements require a calibrated radiometric chain or integrating geometry and corrections appropriate to the specimen. A relative spectrum can still be highly useful, but it should not be labeled an absolute quantum yield.

Spectral calibration has wavelength, intensity, and line-shape dimensions. Wavelength standards constrain the energy axis; a calibrated source or detector transfer function corrects spectral sensitivity; a narrow reference feature measures instrument broadening. Detector dark signal, cosmic events, grating-order leakage, saturation, polarization response, slit width, and stitching between detector ranges can all reshape a spectrum. Baseline subtraction and smoothing must preserve weak defect bands and peak areas, and raw data should remain available so alternate physically justified fits can be tested.

Steady-state PL and time-resolved PL answer related but different questions. Steady-state spectra reveal the occupied radiative pathways under a maintained generation condition. Time-resolved photoluminescence observes decay after pulsed excitation, but even a decay constant can combine bulk, surface, trapping, diffusion, photon recycling, and instrument-response effects. A steady-state intensity map may correlate with lifetime after calibration for a defined material and injection regime; the correlation is not a universal conversion. Specialized lifetime mapping therefore deserves its own excitation, temporal-response, and transport model rather than being silently inferred here.

A production PL result becomes defensible when it states what was generated, which pathways competed, how emitted photons were transferred to the detector, and which reference makes the inference quantitative. That is the generation-recombination-and-optical-transfer lens.