Carrier generation and recombination are the processes that create and remove mobile charge carriers in a semiconductor. Generation commonly promotes an electron from the valence band to the conduction band, leaving an electron hole behind; recombination reverses this change by transferring a conduction-band electron into an available valence-band state. These processes alter the population of mobile carriers without creating or destroying net electric charge. They are distinct from the creation and annihilation of elementary particles.
Generation and recombination determine how a semiconductor responds to illumination, electrical injection, temperature, and energetic radiation. They control the conversion of light into electricity in solar cells and photodetectors, the production of light in light-emitting diodes and lasers, and the switching and leakage characteristics of electronic devices. Their rates depend on electronic structure, carrier populations, crystal defects, interfaces, and the exchange of energy with photons, lattice vibrations, or other carriers.
A complete description combines microscopic transition mechanisms with carrier transport and statistical physics. Radiative, defect-assisted, and Auger recombination can operate simultaneously, while thermal excitation, optical absorption, impact ionization, and field-assisted tunneling supply carriers. Their competition determines carrier lifetime, diffusion length, luminescence efficiency, and the departure of an operating device from thermal equilibrium.
Carrier populations and thermal equilibrium
The distinction between generation and recombination follows from a semiconductor’s electronic band structure. The valence band contains the highest occupied electronic states at zero temperature in an ideal intrinsic semiconductor, while the conduction band provides higher-energy states in which electrons can move through the crystal. The intervening band gap, , sets a principal energy scale for interband excitation. At approximately 300 K, crystalline silicon has a gap near , while gallium arsenide has a gap near . Both the magnitude of the gap and the locations of the band extrema in crystal momentum strongly influence transition rates.
Let denote the conduction-electron concentration and the hole concentration. For a nondegenerate semiconductor in thermal equilibrium,
where and are effective band densities of states, and are band-edge energies, is the equilibrium Fermi level, is the Boltzmann constant, and is absolute temperature. These expressions are Boltzmann approximations to Fermi–Dirac statistics. Multiplication gives
where is the intrinsic carrier concentration. Strong degeneracy, heavy-doping effects, or significant modification of the band structure requires a more complete treatment.
Thermal equilibrium does not imply that generation and recombination stop. Instead, their average rates balance. If and are volumetric generation and recombination rates, both commonly expressed in , equilibrium requires . Microscopic reversibility relates each elementary transition to its inverse. Consequently, a material can have rapid microscopic carrier exchange while maintaining constant macroscopic carrier populations. The equilibrium condition must include the relevant thermal environment, including its radiation field when radiative transitions are considered.
Outside equilibrium, the quantity is a useful measure of the tendency toward net recombination in many standard models. Optical excitation or electrical injection usually makes it positive; extraction from a depleted region can make it negative, so that thermal generation dominates. When carriers within each band have approximately thermalized, separate quasi-Fermi levels and describe their populations. In the nondegenerate limit,
The quasi-Fermi-level separation therefore connects nonequilibrium carrier populations to the voltage and light-emission characteristics of semiconductor devices.
Mechanisms of carrier generation
Optical generation begins when a semiconductor absorbs a photon and transfers its energy to the electronic system. An interband transition generally requires photon energy at least comparable to the band gap, although excitonic states, disorder, defects, and strong fields can modify the absorption edge. For monochromatic illumination entering a homogeneous material at , a simple single-pass model gives
where is the absorption coefficient, is incident photon flux, is front-surface reflectance, and is the number of initially generated mobile electron–hole pairs per absorbed photon. Interference, light trapping, spatially varying absorption, and multiple reflections require a more detailed optical calculation.
In a direct-gap semiconductor, the conduction-band minimum and valence-band maximum occur at the same crystal momentum. Because an optical photon carries little momentum on the scale of the crystal’s reciprocal lattice, transitions can occur without substantial additional momentum exchange. An indirect-gap semiconductor, such as silicon, generally requires a phonon to conserve momentum in transitions near the absorption edge. This requirement weakens near-edge absorption and makes it more temperature dependent. At higher photon energies, transitions elsewhere in the band structure can produce much stronger absorption.
Photon absorption does not always produce independently mobile carriers immediately. In materials with appreciable electron–hole binding, the initial excitation may be an exciton, a bound electron–hole pair. Its dissociation can depend on temperature, electric field, or an interface between different materials. Conversely, absorption well above the gap creates energetic carriers that typically lose excess energy through carrier–carrier and carrier–phonon interactions. This cooling usually precedes the eventual recombination of the pair, so the absorbed photon energy is not generally retained entirely as electrical free energy.
Thermal generation is the reverse of thermally allowed recombination pathways. Electrons can be excited across the gap directly or through intermediate defect states. The strong temperature dependence of explains why thermally generated dark current becomes especially important in narrow-gap infrared detectors. Ionization of shallow dopants also changes the mobile-carrier population, but it is conceptually distinct from generating an electron–hole pair: donor ionization releases an electron while leaving a positively charged donor, and acceptor ionization produces a hole and a negatively charged acceptor.
Energetic carriers can create additional pairs through impact ionization, transferring sufficient kinetic energy to a valence electron. Repeated events produce avalanche multiplication in sufficiently strong electric fields. Another mechanism is band-to-band quantum tunneling, in which a strong field permits transitions between occupied valence-band states and empty conduction-band states across a spatially narrow forbidden region. Ionizing radiation likewise generates carriers through cascades of excitation and ionization. In silicon near room temperature, the mean energy required to form a collected electron–hole pair under energetic-particle excitation is approximately , greater than the band gap because part of the deposited energy goes into lattice excitation.
Radiative recombination and optical reciprocity
Radiative recombination transfers the energy released by an electron–hole transition to a photon. In a direct-gap material, the simplest nondegenerate, spatially homogeneous model writes the net radiative recombination rate as
where is the bimolecular radiative recombination coefficient. Its common unit is . The gross spontaneous emission rate is approximately proportional to ; the subtraction of the equilibrium term accounts for the reverse thermal optical process. The value of depends on material, temperature, band structure, and, outside the simplest regime, carrier density. It is not a universal constant.
Whether a transition efficiently emits light depends on more than energy conservation. Momentum conservation and optical selection rules determine which electronic states can couple through the electromagnetic field. Direct-gap compounds such as gallium arsenide are effective light emitters because band-edge transitions can satisfy these conditions without a phonon. In silicon, near-edge radiative recombination is phonon assisted and much less probable. Silicon can nevertheless emit measurable light, especially when competing nonradiative processes are suppressed; its indirect gap makes it inefficient as an ordinary light source rather than absolutely nonluminescent.
At low excitation in a p-type material, the hole population remains approximately fixed at its equilibrium value , and the excess electron concentration is . The radiative rate then becomes approximately , giving a minority-carrier radiative lifetime
Under high-level injection, when greatly exceeds the equilibrium carrier populations, radiative recombination instead scales approximately as . Thus, an apparently first-order decay in one regime and second-order decay in another can arise from the same microscopic mechanism.
Spontaneous emission occurs without an externally supplied optical mode population, whereas stimulated emission is enhanced by photons already occupying the relevant mode. Semiconductor lasers exploit stimulated emission under conditions in which injected carrier populations provide optical gain. In a simple quasi-equilibrium description, gain at photon energy requires an appropriate population inversion, associated with a quasi-Fermi-level separation exceeding that photon energy. The actual threshold also depends on optical confinement, cavity loss, and nonradiative carrier loss.
Absorption and emission are constrained by detailed balance. The van Roosbroeck–Shockley relation, introduced in 1954, connects a semiconductor’s equilibrium radiative recombination to its optical absorption and the thermal photon distribution. In operating devices, emitted photons may escape, be parasitically absorbed, or be reabsorbed in the semiconductor to regenerate carriers. This last process, photon recycling, can lengthen the observed carrier decay and improve attainable voltage in high-quality photovoltaic materials. Internal radiative efficiency and external light-emission efficiency therefore differ: the latter also includes optical escape and absorption losses.
Defect-assisted recombination and carrier trapping
Crystal defects introduce electronic states within the band gap that can mediate recombination in successive steps. A defect may first capture a conduction electron and later capture a hole, restoring its initial charge state while removing an electron–hole pair. The released energy commonly enters lattice vibrations through multiphonon processes. Defects that participate include impurities, vacancies, interstitials, complexes, and states associated with extended structural imperfections. Their recombination activity depends on energy level, charge state, capture kinetics, and spatial environment, not merely on their concentration.
The standard single-level Shockley–Read–Hall model was developed independently by William Shockley and William T. Read Jr., and by Robert N. Hall, in 1952. For a spatially uniform population of independent defects, its net recombination rate is
For a simple trap without additional degeneracy factors,
where is the trap energy and is the intrinsic Fermi level. The characteristic parameters are
with defect density , electron and hole capture cross sections , and thermal velocities .
A level near midgap is often an effective recombination or generation center because it can exchange both electrons and holes without one emission process overwhelmingly dominating. This familiar rule is conditional: strongly unequal capture cross sections can shift the most effective energy away from the middle of the gap. Likewise, a high density of defects need not imply a proportionally high recombination rate if those states couple weakly to one carrier type or lie in a region where that carrier is scarce. In a depleted junction, where both mobile-carrier populations are low, the same trap cycle can run in the generation direction and contribute to leakage current.
Trapping and recombination must be distinguished. Capture of an electron temporarily removes it from the conduction band, but if that electron is later emitted back without a hole being captured, no complete electron–hole recombination has occurred. Trapping can nevertheless alter conductivity, delay charge collection, and produce nonexponential transients. Measurements that infer carrier lifetime from electrical or optical decay can therefore be affected by redistribution between free and trapped populations.
Real defects frequently exceed the assumptions of the single-level model. A center may possess multiple charge states, undergo a structural rearrangement after capture, or interact with nearby defects. In disordered semiconductors, a distribution of gap states replaces one discrete energy. Strong fields can enhance emission or enable tunneling-assisted capture. These cases require coupled occupancy equations, distributions of transition rates, or microscopic calculations rather than an unqualified use of a single Shockley–Read–Hall lifetime.
Auger recombination and density-dependent losses
In Auger recombination, an electron and a hole recombine while transferring their energy to another carrier rather than to a photon. The recipient may be an electron or a hole, which becomes highly energetic and subsequently relaxes through scattering. Energy and momentum must be conserved across the entire event. Auger recombination is therefore an intrinsic nonradiative process: it can occur even in an otherwise perfect crystal and does not require a deep defect.
A widely used nondegenerate expression is
where and , commonly expressed in , describe electron-assisted and hole-assisted channels. Far from equilibrium, with the equilibrium contribution negligible, the respective terms scale as and . This dependence makes Auger recombination increasingly important at high injected carrier density or high doping. The coefficients can depend on temperature and carrier concentration because screening, degeneracy, and band-structure effects modify the available transitions.
For approximately equal electron and hole populations , the Auger rate is roughly proportional to . The corresponding effective lifetime varies approximately as . In a heavily doped p-type region under low injection, by contrast, hole-assisted Auger recombination gives an excess-electron lifetime approximately proportional to . A process that is third order under high injection can consequently appear first order in a small excess minority-carrier population.
Auger channels are sensitive to material properties. A narrow gap reduces the energy that must be transferred, often making Auger loss particularly important in infrared semiconductors. Band offsets, strain, and the arrangement of conduction and valence subbands can open or suppress particular channels in quantum wells. Phonons and alloy disorder can assist momentum conservation. In strongly confined nanocrystals, Coulomb coupling and modified momentum constraints can make multicarrier Auger decay especially rapid; such behavior cannot always be represented by bulk coefficients.
The inverse process is impact ionization: an energetic carrier loses energy by creating an additional electron–hole pair. Ordinary thermalized device operation generally favors net Auger loss, whereas strong fields or energetic optical excitation can drive the reverse process. This relation connects high-density recombination with avalanche multiplication without making the two interchangeable descriptions of a single operating regime.
Auger recombination is one possible contributor to the decrease of light-emitting-diode efficiency at high current density, commonly called efficiency droop. Carrier leakage, imperfect injection, current crowding, heating, and defect-related mechanisms can also contribute. A cubic term fitted to light-output data does not by itself establish a unique microscopic Auger mechanism, because the relationship between injected current, active-region carrier density, and measured emission can be complicated.
Surfaces, interfaces, and spatially localized recombination
A semiconductor surface interrupts the periodic crystal lattice and can leave unsatisfied chemical bonds or other localized electronic states. Interfaces with oxides, metals, or another semiconductor introduce additional structural and chemical environments. These regions may have far higher recombination activity per unit volume than the bulk. As device dimensions shrink, or when thin films and nanostructures are used, their large surface-to-volume ratio makes interface behavior particularly important.
A common simplified boundary condition characterizes a surface by a surface recombination velocity . Under low injection and an approximately fixed majority-carrier population, the excess minority-carrier recombination flux per unit area is
where is the excess minority-carrier concentration at the surface. The parameter has units of velocity, usually , but it is not the actual directed speed of individual carriers. It summarizes how effectively arriving carriers are removed. More complete interface models use both carrier concentrations, trap-energy distributions, capture parameters, and electrostatic band bending.
For a planar sample of thickness , with nearly uniform excess carrier density and two comparable surfaces, a useful approximation is
This expression shows why a thin specimen can exhibit a short measured lifetime despite an excellent bulk crystal. It assumes that diffusion maintains a sufficiently uniform profile. If a strongly recombining surface creates a pronounced carrier gradient, a diffusion boundary-value problem must replace the uniform-density approximation. The resulting decay can contain several spatial modes rather than a single exponential.
Surface passivation suppresses recombination through two principal mechanisms. Chemical passivation reduces the number or activity of electrically harmful interface states, for example by satisfying dangling bonds. Field-effect passivation uses fixed charge or an electrostatic potential to reduce the concentration of one carrier species at the interface. Since recombination requires access to both electrons and holes, excluding either species can lower the rate even when some interface states remain. Silicon dioxide, hydrogen-containing silicon nitride, and aluminum oxide are important passivation materials, with performance depending on processing and substrate conditions.
Heterointerfaces introduce band offsets that can separate electrons and holes or selectively transmit one carrier type. This is useful in photovoltaic contacts and light-emitting structures, but unintended offsets can also cause carrier accumulation and enhance local recombination. Grain boundaries and dislocations similarly combine structural defects with possible electrostatic barriers. Their activity varies with atomic structure, impurity decoration, and charge state. It is therefore insufficient to classify every grain boundary as either harmless or strongly recombining without specifying its electronic and chemical properties.
Continuity equations, lifetime, and diffusion length
Microscopic rates enter device physics through carrier conservation. Using conventional electron and hole current densities and , the continuity equations are
where is the positive elementary charge. The opposite divergence signs follow from the charges carried by electrons and holes. For direct interband pair generation or recombination, the electron and hole source terms are equal. During defect capture and emission, however, the mobile populations can change at different instantaneous rates because charge is temporarily stored on the defects.
Transport couples these equations to the electric field and carrier-concentration gradients. In a simple drift–diffusion description,
Here are mobilities and are diffusion coefficients. The field is obtained self-consistently from the electrostatic charge distribution. Consequently, an observed reduction in local carrier concentration can reflect transport out of the measurement region, recombination there, or both. Charge collection at an electrode must not automatically be interpreted as a bulk recombination event.
For a homogeneous excess population, a commonly defined effective lifetime is
when electrons and holes are generated in equal excess and is the net recombination rate removing that excess. If with constant , turning off generation gives exponential decay. Nonlinear radiative and Auger terms instead produce density-dependent decay. Under maintained steady excitation, a small perturbation probes a differential lifetime,
which need not equal .
For independent channels evaluated at the same carrier populations, rates add. Defining each channel’s effective lifetime consistently then gives
This reciprocal rule is not permission to combine lifetimes measured under different injection levels or boundary conditions. Nor does it fully represent coupled trapping, spatial transport, or photon recycling without additional terms.
The carrier diffusion length in a simple linear regime is
It is the characteristic distance over which an excess-carrier distribution decays when diffusion competes with recombination. Minority-carrier collection in a solar cell often depends on whether this distance is comparable to the distance to a collecting junction. Where electrons and holes move together while preserving approximate charge neutrality, an ambipolar diffusion coefficient is appropriate. Strong electric fields, nonuniform material properties, and nonlinear recombination require a spatial transport calculation rather than a single universal collection length.
Consequences for electronic and optoelectronic devices
A p–n junction separates regions with different majority carriers and establishes a depletion region with an internal electric field. Under forward bias, injected minority carriers diffuse away from the junction and recombine. In the ideal low-injection diffusion model,
The saturation-current density depends on equilibrium minority-carrier populations, transport coefficients, and recombination. Recombination through suitable depletion-region defects can instead produce an approximate voltage dependence . The associated ideality factors near 1 and 2 describe restricted regimes, not unique labels for all real-device mechanisms; series resistance, tunneling, high injection, and nonuniformity can change the measured slope.
In a solar cell, optical generation supplies carriers while selective contacts extract them as electrical current. Recombination removes carriers before collection and also limits voltage. At open circuit, there is no net current through the external circuit, so the device reaches a steady state in which total generation is balanced by total recombination. Suppressing nonradiative channels allows a larger quasi-Fermi-level separation for the same illumination. Radiative recombination remains even in an ideal absorber, establishing the detailed-balance constraint underlying the Shockley–Queisser treatment of single-junction photovoltaic conversion.
A light-emitting diode uses electrical injection to drive radiative recombination. Its internal quantum efficiency is related to the radiative share of total carrier recombination, but overall performance also depends on whether injected carriers reach the active region. External quantum efficiency additionally includes the probability that generated photons escape. A semiconductor laser must sustain enough carrier density for optical gain to balance cavity losses. Recombination below and near threshold determines much of the required injection current, while stimulated emission becomes a major carrier-removal process above threshold.
In a photodetector, generation is the desired signal process, but thermally driven generation produces dark current. Depletion-region defects can generate carriers even without illumination, while surfaces and contacts can add leakage. Avalanche photodiodes use impact ionization to amplify an initial signal, introducing statistical multiplication noise. Photoconductors can achieve gain when a carrier lifetime is long compared with transit time, but the same long-lived population can slow response. Trapping may produce persistent photoconductivity, so high gain does not necessarily imply rapid or faithful temporal detection.
Power devices present a different trade-off. Long minority-carrier lifetime can increase conductivity through injected carrier populations and reduce conduction loss, but stored charge must be removed during turn-off. Controlled introduction of recombination centers or irradiation-induced defects can shorten lifetime and reduce switching tails, at the cost of altered conduction and leakage behavior. In memory and logic structures, unintended generation contributes to leakage and charge loss. Device engineering therefore does not seek the longest possible lifetime in every case; it tailors the spatial distribution and operating dependence of generation and recombination to the function of the component.
Experimental methods, historical development, and interpretation
Carrier lifetime measurements infer recombination from an observable linked to carrier population. In time-resolved photoluminescence, a light pulse generates carriers and the emitted intensity is measured as a function of delay. Interpretation requires a model because luminescence generally follows , not carrier concentration directly. Under high injection with , intensity is approximately proportional to the square of excess density; under low injection in a doped material, it can be approximately linear. Carrier diffusion, optical reabsorption, and the instrument response further shape the recorded transient.
Photoconductance methods measure the electrical conductivity change caused by illumination. For approximately equal excess electron and hole populations,
Transient photoconductance and microwave photoconductance decay track the return toward equilibrium after excitation. Quasi-steady-state methods estimate recombination from the balance between known generation and measured excess carrier density, with a correction when that density changes appreciably during the measurement. Conversion from conductivity to carrier concentration requires suitable mobility values. Surface preparation, sample thickness, temperature, and injection level must therefore accompany a meaningful lifetime result.
Defect-specific measurements supply complementary information. Deep-level transient spectroscopy, introduced by D. V. Lang in 1974, measures capacitance transients associated with thermally activated carrier emission from defects in junction structures. Temperature-dependent emission can reveal activation energies and capture-related parameters. Electron-beam-induced current maps how locally generated carriers contribute to collected current, exposing recombination-active boundaries and defects. Spatially resolved photoluminescence and cathodoluminescence similarly reveal inhomogeneity, although intensity contrast may reflect absorption, carrier transport, or optical extraction as well as recombination.
The theoretical framework emerged alongside early transistor and semiconductor-device physics. The 1952 Shockley–Read and Hall papers established a quantitative description of capture and emission through gap states. The 1954 van Roosbroeck–Shockley relation connected radiative recombination with optical absorption. Subsequent work extended these descriptions to heavy doping, indirect transitions, strong fields, heterostructures, and low-dimensional systems. Modern atomistic calculations use electronic structure and lattice relaxation to predict defect energies and nonradiative capture, while device simulations embed these microscopic parameters in transport equations.
A persistent interpretive problem is that different mechanisms can produce similar measured dependences. The frequently used high-injection model
associates the three terms with defect-assisted, radiative, and Auger recombination under specific assumptions. Doping, trap occupancy, carrier leakage, spatial nonuniformity, and density-dependent coefficients can invalidate those simple identifications. Reliable mechanism assignment therefore combines multiple observables, such as injection-dependent lifetime, temperature dependence, absolute luminescence efficiency, defect spectroscopy, and spatially resolved transport. In emerging thin-film and nanostructured materials, coupled exciton dynamics, interfacial charge transfer, mobile ionic defects, and strongly varying local environments make this distinction between a fitted rate law and a demonstrated microscopic process especially important.