Photoluminescence (PL)

Photoluminescence is emission of light following optical excitation. In a semiconductor, a pump photon with energy at or above the bandgap creates an electron–hole pair; the carriers relax, mostly by phonon emission, to the band edges within picoseconds and then recombine, some of them radiatively. The emitted spectrum sits near the band edge: bulk GaAs peaks near 1.424 eV (871 nm) at 300 K and near 1.52 eV (816 nm) at liquid-helium temperature, InP near 1.344 eV (923 nm) at 300 K. Because it needs no contacts and does not damage the sample, PL is the quick-look inspection method for epitaxial wafers.

Measurement setup. A standard PL system:

Detectors are matched to the emission: silicon arrays below about 1 μm, InGaAs arrays out to 1.7 μm. For weak signals the pump is chopped and the detector output is read with a lock-in amplifier. Wafer mappers raster the pump spot and record peak wavelength and intensity at each point, which maps composition, quantum well thickness, and defect density across a wafer after epitaxy.

Spectral features. PL spectra of III–V semiconductors typically show:

• Band-edge emission at the fundamental bandgap (or at the quantum-confined energy in a quantum well)
• Defect-related emission at sub-bandgap energies: broad features from deep defect levels
• Phonon replicas: discrete peaks shifted below the band edge by optical phonon energies (LO phonon $\approx$ 36 meV for GaAs)
• Exciton emission at low temperature: bound electron–hole pairs producing sharp features just below the band gap

Lineshape and temperature. For band-to-band emission from a bulk direct-gap semiconductor with nondegenerate carriers, the spectrum follows

$$I(E)\propto \sqrt{E-E_g}{e}^{-(E-E_g)\mathrm{/}kT}\mathrm{.}$$

It peaks at $E_g+kT\mathrm{/}2$ and has a full width of about $1.8kT$: 46 meV at 300 K, where $kT=25.9$ meV. At 4 K, $kT=0.34$ meV, which is why cryogenic PL resolves excitonic and shallow-impurity lines a few meV apart that merge at room temperature.

Excitation density. A 1 mW pump at 532 nm delivers $2.7\times {10}^{15}$ photons/s. Focused to a 10 μm diameter spot, that is a flux of $3.4\times {10}^{21}$ cm⁻² s⁻¹. If all photons are absorbed within 1 μm and the carrier lifetime is 1 ns, the steady-state carrier density is about $3.4\times {10}^{16}$ cm⁻³, neglecting diffusion. Pump power sets the injection regime, so it should be quoted with every PL result.

PL intensity versus pump power. Integrated PL is fitted to $I_{\text{PL}}\propto P^k$ on a log-log plot. Excitonic emission gives $k\approx 1$. Band-to-band emission gives $k\approx 2$ when non-radiative recombination through defects controls the carrier density, because the radiative rate scales as $np$; it falls toward 1 as radiative recombination takes over. At very high pump power $k<1$ from saturation of carrier capture or Auger recombination.

Time-resolved PL (TRPL). After a short pump pulse, the PL decays with lifetime

$$\frac{1}{{\tau }_{PL}}=\frac{1}{{\tau }_{rad}}+\frac{1}{{\tau }_{nr}},$$

where ${\tau }_{rad}$ and ${\tau }_{nr}$ are the radiative and non-radiative lifetimes. The internal quantum efficiency is $\eta ={\tau }_{PL}\mathrm{/}{\tau }_{rad}$. With ${\tau }_{rad}=10$ ns and ${\tau }_{nr}=2$ ns, for example, ${\tau }_{PL}=1.67$ ns and $\eta =17\mathrm{\%}$. Comparing ${\tau }_{PL}$ at room and cryogenic temperature separates the non-radiative contribution, since non-radiative recombination through defects is thermally activated. Good III–V epilayers typically show lifetimes of a nanosecond or longer at room temperature; lifetimes of tens of picoseconds indicate heavy defect recombination. TRPL is recorded with time-correlated single-photon counting or a streak camera.

Quantum yield. The PL quantum yield is emitted photons per absorbed photon. Measured in power, it must be corrected by the photon-energy ratio: a sample pumped at 532 nm (2.33 eV) and emitting at 1550 nm returns at most 34% of the absorbed power even at unit quantum yield.

Pitfalls. Only a small fraction of internally generated light escapes a high-index sample: for $n=3.5$ the escape cone through one flat face holds about 2% of isotropic emission, before Fresnel loss. Absolute PL intensities therefore compare samples only when geometry, surface, and collection optics are identical. Reabsorption in thick samples redshifts the observed peak. Pump heating shifts the peak to longer wavelength as $E_g$ drops, which can be mistaken for a composition change. Indirect-gap materials such as silicon (1.12 eV, 1107 nm) emit weakly because radiative recombination requires a phonon; see direct versus indirect bandgap.

Relation to other techniques. Electroluminescence (EL) is the analogous measurement under electrical injection and is used once contacts are fabricated; PL replaces electrical injection with optical pumping.

Common questions

What is the difference between photoluminescence and fluorescence?

Fluorescence is photoluminescence from a spin-allowed transition with a short lifetime, typically nanoseconds, in molecules and dyes. Phosphorescence is photoluminescence from a spin-forbidden transition, lasting microseconds to hours. In semiconductor work, "PL" covers all optically excited emission.

What does a photoluminescence spectrum show?

The peak energy gives the bandgap or quantum-well transition energy, and hence composition or well width. Linewidth reflects temperature and disorder. Sub-gap bands indicate defects. Integrated intensity and decay time indicate the balance between radiative and non-radiative recombination.

References: Pankove, Optical Processes in Semiconductors (Prentice-Hall, 1971; Dover, 1975); Yu & Cardona, Fundamentals of Semiconductors (4th ed., 2010); Schroder, Semiconductor Material and Device Characterization (3rd ed., 2006); Saleh & Teich, Fundamentals of Photonics (3rd ed., 2019).