In a typical PL experiment, a semiconductor is excited with a light-source that provides photons with an energy larger than the
bandgap energy. The incoming light excites a polarization that can be described with the
semiconductor Bloch equations. Once the photons are absorbed, electrons and holes are formed with finite momenta \mathbf{k} in the
conduction and
valence bands, respectively. The excitations then undergo energy and momentum relaxation towards the band-gap minimum. Typical mechanisms are
Coulomb scattering and the interaction with
phonons. Finally, the electrons recombine with holes under emission of photons. Ideal, defect-free semiconductors are
many-body systems where the interactions of charge-carriers and lattice vibrations have to be considered in addition to the light-matter coupling. In general, the PL properties are also extremely sensitive to internal
electric fields and to the dielectric environment (such as in
photonic crystals) which impose further degrees of complexity. A precise microscopic description is provided by the
semiconductor luminescence equations. and Balkan (1998). The fictive model structure for this discussion has two confined quantized electronic and two hole
subbands, e1, e2 and h1, h2, respectively. The linear
absorption spectrum of such a structure shows the
exciton resonances of the first (e1h1) and the second quantum well subbands (e2, h2), as well as the absorption from the corresponding continuum states and from the barrier.
Photoexcitation In general, three different excitation conditions are distinguished: resonant, quasi-resonant, and non-resonant. For the resonant excitation, the central energy of the laser corresponds to the lowest
exciton resonance of the
quantum well. No, or only a negligible amount of the excess, energy is injected to the carrier system. For these conditions, coherent processes contribute significantly to the spontaneous emission. The decay of polarization creates excitons directly. The detection of PL is challenging for resonant excitation as it is difficult to discriminate contributions from the excitation, i.e., stray-light and diffuse scattering from surface roughness. Thus,
speckle and resonant
Rayleigh-scattering are always superimposed to the
incoherent emission. In case of the non-resonant excitation, the structure is excited with some excess energy. This is the typical situation used in most PL experiments as the excitation energy can be discriminated using a
spectrometer or an
optical filter. One has to distinguish between quasi-resonant excitation and barrier excitation. For quasi-resonant conditions, the energy of the excitation is tuned above the ground state but still below the
barrier absorption edge, for example, into the continuum of the first subband. The polarization decay for these conditions is much faster than for resonant excitation and coherent contributions to the quantum well emission are negligible. The initial temperature of the carrier system is significantly higher than the lattice temperature due to the surplus energy of the injected carriers. Finally, only the electron-hole plasma is initially created. It is then followed by the formation of excitons. In case of barrier excitation, the initial carrier distribution in the quantum well strongly depends on the carrier scattering between barrier and the well.
Relaxation Initially, the laser light induces coherent polarization in the sample, i.e., the transitions between electron and hole states oscillate with the laser frequency and a fixed phase. The polarization dephases typically on a sub-100 fs time-scale in case of nonresonant excitation due to ultra-fast Coulomb- and phonon-scattering. The dephasing of the polarization leads to creation of populations of electrons and holes in the conduction and the valence bands, respectively. The lifetime of the carrier populations is rather long, limited by radiative and non-radiative recombination such as
Auger recombination. During this lifetime a fraction of electrons and holes may form excitons, this topic is still controversially discussed in the literature. The formation rate depends on the experimental conditions such as lattice temperature, excitation density, as well as on the general material parameters, e.g., the strength of the Coulomb-interaction or the exciton binding energy. The characteristic time-scales are in the range of hundreds of
picoseconds in GaAs; Directly after the excitation with short (femtosecond) pulses and the quasi-instantaneous decay of the polarization, the carrier distribution is mainly determined by the spectral width of the excitation, e.g., a
laser pulse. The distribution is thus highly non-thermal and resembles a
Gaussian distribution, centered at a finite momentum. In the first hundreds of
femtoseconds, the carriers are scattered by phonons, or at elevated carrier densities via Coulomb-interaction. The carrier system successively relaxes to the
Fermi–Dirac distribution typically within the first picosecond. Finally, the carrier system cools down under the emission of phonons. This can take up to several
nanoseconds, depending on the material system, the lattice temperature, and the excitation conditions such as the surplus energy. Initially, the carrier temperature decreases fast via emission of
optical phonons. This is quite efficient due to the comparatively large energy associated with optical phonons, (36meV or 420K in GaAs) and their rather flat dispersion, allowing for a wide range of scattering processes under conservation of energy and momentum. Once the carrier temperature decreases below the value corresponding to the optical phonon energy,
acoustic phonons dominate the relaxation. Here, cooling is less efficient due their
dispersion and small energies and the temperature decreases much slower beyond the first tens of picoseconds. At elevated excitation densities, the carrier cooling is further inhibited by the so-called
hot-phonon effect. The relaxation of a large number of hot carriers leads to a high generation rate of optical phonons which exceeds the decay rate into acoustic phonons. This creates a non-equilibrium "over-population" of optical phonons and thus causes their increased reabsorption by the charge-carriers significantly suppressing any cooling. Thus, a system cools slower, the higher the carrier density is.
Radiative recombination The emission directly after the excitation is spectrally very broad, yet still centered in the vicinity of the strongest exciton resonance. As the carrier distribution relaxes and cools, the width of the PL peak decreases and the emission energy shifts to match the ground state of the exciton (such as an electron) for ideal samples without disorder. The PL spectrum approaches its quasi-steady-state shape defined by the distribution of electrons and holes. Increasing the excitation density will change the emission spectra. They are dominated by the excitonic ground state for low densities. Additional peaks from higher subband transitions appear as the carrier density or lattice temperature are increased as these states get more and more populated. Also, the width of the main PL peak increases significantly with rising excitation due to excitation-induced dephasing and the emission peak experiences a small shift in energy due to the Coulomb-renormalization and phase-filling. in the lattice or
disorder due to variations of the chemical composition. Their treatment is extremely challenging for microscopic theories due to the lack of detailed knowledge about perturbations of the ideal structure. Thus, the influence of the extrinsic effects on the PL is usually addressed phenomenologically. In experiments, disorder can lead to localization of carriers and hence drastically increase the photoluminescence life times as localized carriers cannot as easily find nonradiative recombination centers as can free ones. Researchers from the
King Abdullah University of Science and Technology (KAUST) have studied the photoinduced
entropy (i.e. thermodynamic disorder) of
InGaN/
GaN p-i-n
double-heterostructure and
AlGaN nanowires using temperature-dependent photoluminescence. They defined the photoinduced
entropy as a thermodynamic quantity that represents the unavailability of a system's energy for conversion into useful work due to
carrier recombination and
photon emission. They have also related the change in entropy generation to the change in photocarrier dynamics in the nanowire active regions using results from time-resolved photoluminescence study. They hypothesized that the amount of generated disorder in the
InGaN layers eventually increases as the temperature approaches room temperature because of the thermal activation of
surface states, while an insignificant increase was observed in AlGaN nanowires, indicating lower degrees of disorder-induced uncertainty in the wider bandgap semiconductor. To study the photoinduced
entropy, the scientists have developed a mathematical model that considers the net energy exchange resulting from photoexcitation and photoluminescence. == Photoluminescence from metals ==