A photodiode is a light-sensitive semiconductor device in which absorbed optical radiation generates mobile charge carriers that produce an electrical signal. Most photodiodes use a p–n junction or a closely related junction structure to separate these carriers. The resulting photocurrent provides a measure of the incident light, subject to the device’s wavelength response, collection efficiency, and operating conditions. Photodiodes detect radiation from the ultraviolet through the infrared, with the accessible spectral range determined primarily by the semiconductor material and device structure.
Photodiodes are central components of optical receivers, scientific instruments, industrial sensors, and electronic imaging systems. They can measure steady illumination, resolve rapidly varying optical signals, or, in specialized forms, register individual photons. Their performance involves trade-offs among sensitivity, speed, noise, active area, spectral coverage, and electrical power consumption. Although closely related to solar cells and other light-sensitive semiconductor devices, photodiodes are generally optimized for detecting or measuring radiation rather than delivering substantial electrical power.
Physical principles and carrier collection
The operation of a photodiode begins with the absorption of a photon. In an ordinary semiconductor interband transition, the photon transfers enough energy to promote an electron from the valence band into the conduction band, leaving a mobile hole behind. The minimum energy required is set approximately by the material’s band gap, . Photon energy is related to wavelength by
where is the Planck constant, is optical frequency, is the speed of light, and is wavelength. The approximate long-wavelength absorption threshold is therefore , although practical detector response also depends on absorption strength, temperature, and device construction.
A junction is formed by bringing together semiconductor regions with different electrical properties, commonly through semiconductor doping. In a p–n junction, diffusion of electrons and holes across the interface leaves behind fixed ionized dopants. These charges establish a built-in electric field in a region depleted of mobile carriers, called the depletion region. An electron–hole pair generated within this region is separated by the field: electrons move toward the n side and holes toward the p side. Their motion produces current in the external circuit.
Photons absorbed outside the depletion region can also contribute to the signal. Carriers generated in electrically neutral material may diffuse toward the junction and then be swept across it. Their contribution depends on the distance to the junction and on carrier lifetime, which determines how long they survive before recombination. Diffusion collection can improve total sensitivity but may introduce a slow tail in the electrical response. Carriers lost to recombination at defects, interfaces, or surfaces do not contribute fully to the measured signal.
At the microscopic level, absorption and charge collection are distinct processes. A photon may be reflected before entering the device, transmitted without being absorbed, or absorbed in a region from which carriers cannot be collected efficiently. Antireflection coatings, suitable absorption-layer thicknesses, surface passivation, and electric-field engineering address these different losses. A high-performance photodiode must therefore combine effective optical coupling with efficient electrical collection; a material that absorbs strongly is not automatically an efficient detector.
Electrical operation and bias modes
An illuminated photodiode can be represented approximately as a diode in parallel with a light-generated current source, together with junction capacitance, shunt resistance, and series resistance. Using a conventional forward-diode current direction, a simplified current–voltage relationship is
where is saturation current, is the magnitude of elementary electric charge, is junction voltage, is an ideality factor, is the Boltzmann constant, is absolute temperature, and is photocurrent. Real devices depart from this expression because of leakage paths, recombination, series resistance, and field-dependent processes.
In photovoltaic operation, no external reverse-bias supply is applied. With an open circuit, illumination produces a photovoltage; with a short circuit or a suitable amplifier maintaining nearly zero terminal voltage, it produces a measurable current. The latter arrangement is common in precision optical measurements because it avoids the increase in leakage associated with reverse bias. A photodiode connected to a transimpedance amplifier can thus operate near zero junction voltage while still producing a useful output voltage. Zero bias does not mean zero signal or zero built-in electric field.
In photoconductive operation, the junction is reverse biased. Reverse bias usually widens the depletion region, lowers junction capacitance, and strengthens carrier collection. These changes can increase bandwidth and reduce slow diffusion contributions. Once collection is essentially complete, however, raising the reverse voltage does not necessarily increase the primary photocurrent substantially. Its main benefits may instead be faster response and greater tolerance of high optical power. Reverse bias also tends to increase dark current and requires operation within the device’s voltage rating.
The terms “photovoltaic” and “photoconductive” describe operating modes rather than necessarily different physical devices. A junction photodiode may support both modes. This use of “photoconductive” should also be distinguished from a photoconductor, in which light changes the conductivity of a material without requiring a rectifying junction. Phototransistors provide another distinction: they use transistor action to amplify an optically generated signal, often obtaining greater output current at the expense of speed or linearity.
A solar cell uses substantially the same junction physics but is operated at a voltage and current chosen to extract power. A measurement photodiode is instead commonly optimized for calibrated response, low noise, small capacitance, or rapid signal recovery. The boundary is functional rather than a fundamental difference in the photoelectric conversion process.
Junction structures and internal multiplication
The simplest junction photodiode uses a p–n structure. Its depletion width depends on doping and applied voltage, so a substantial fraction of incident light may be absorbed outside the field-filled region. For applications requiring faster and more efficient collection, the p–i–n photodiode places a relatively thick, lightly doped layer between heavily doped p-type and n-type regions. The letter “i” denotes an approximately intrinsic region; practical devices often use material that is lightly doped rather than perfectly intrinsic.
Under appropriate bias, most or all of the central layer is depleted. This provides a comparatively large volume in which light can generate carriers that are collected by drift rather than slow diffusion. Increasing its thickness can improve absorption and lower capacitance for a given area, but carriers then travel farther. The design therefore balances optical absorption against transit time. Waveguide-coupled photodiodes partly separate these requirements by allowing light to travel along the device while carriers cross a much thinner layer.
An avalanche photodiode, or APD, operates with an electric field strong enough for impact ionization. A carrier accelerated by the field can create additional electron–hole pairs, which can themselves produce further carriers. Below breakdown, this process provides finite internal gain, conventionally expressed as a multiplication factor . The average signal current is approximately , but the multiplication process also introduces noise because the number and timing of ionization events fluctuate.
Separate absorption and multiplication structures allow different semiconductor regions to perform the two functions. In an infrared APD, for example, an absorbing layer may be selected for its sensitivity at a telecommunications wavelength, while an adjacent wider-band-gap layer supports high-field multiplication. Additional charge-control layers shape the electric field so that the absorption region does not suffer excessive tunneling or premature breakdown.
A single-photon avalanche diode, or SPAD, is biased above avalanche breakdown. A single initiating carrier can then trigger a self-sustaining avalanche. A quenching circuit must reduce the junction voltage to stop it, after which the detector is recharged. SPADs produce discrete detection events rather than a continuously proportional analog current. Their important characteristics include detection probability, timing jitter, dark-count rate, dead time, and afterpulsing. An avalanche pulse generally does not reveal how many photons arrived simultaneously in that particular sensitive element.
Metal–semiconductor junctions can also form photodiodes. Schottky-barrier devices exploit a rectifying metal contact and may offer useful high-speed or spectral properties. Their absorption and carrier-injection mechanisms differ in detail from those of ordinary p–n devices.
Semiconductor materials and spectral response
Silicon is the dominant photodiode material for visible and near-infrared detection. At room temperature its band gap is approximately , corresponding to a long-wavelength threshold near . Commercial devices commonly cover much of the visible spectrum and part of the near infrared, while specialized construction can extend useful response into the ultraviolet. Silicon benefits from mature manufacturing, relatively low leakage, and compatibility with integrated circuits.
Its wavelength response is not flat. Ultraviolet and blue photons are absorbed close to the surface, making surface recombination, contact geometry, and passivation especially important. Near the long-wavelength limit, absorption is weak and light penetrates much farther. Thick, high-purity silicon can improve collection at these wavelengths, but it complicates fast carrier extraction. Consequently, a silicon detector optimized for ultraviolet measurement need not be optimal for near-infrared timing.
Germanium has a smaller band gap, approximately at room temperature, and extends detection farther into the infrared. Its practical performance depends strongly on device design and temperature; its smaller gap also makes thermally generated leakage more difficult to suppress. Germanium is particularly important in silicon photonics, where germanium absorbing regions can be integrated with silicon waveguides. This combines near-infrared detection with fabrication methods related to silicon electronics.
Indium gallium arsenide, usually abbreviated InGaAs, is widely used around the and wavelengths important to optical communications. A common composition lattice-matched to indium phosphide has a room-temperature cutoff near . Modified compositions extend the spectral range, typically with changes in leakage and fabrication difficulty. Indium phosphide also provides useful multiplication regions and structural layers in compound-semiconductor receivers.
Wide-band-gap materials such as silicon carbide, gallium nitride, and aluminum gallium nitride serve ultraviolet detection. Depending on composition and band gap, they can reject visible light intrinsically rather than relying entirely on an optical filter. This is useful when a weak ultraviolet signal must be distinguished from strong visible illumination.
For longer-wavelength infrared detection, materials such as indium antimonide and mercury cadmium telluride are used. Their narrow gaps allow detection of lower-energy photons but also make thermal carrier generation more significant, so cooling is common in demanding applications. Organic semiconductors, perovskites, and quantum dots provide additional routes to spectral tuning and large-area fabrication, although stability, uniformity, and noise remain important practical considerations.
Responsivity, quantum efficiency, and speed
The basic measure of a photodiode’s optical-to-electrical conversion is responsivity, usually denoted and expressed in amperes per watt:
where is optical power at a specified reference plane. Responsivity must be associated with wavelength, bias, and temperature. It also matters whether the quoted power is incident on the package, on the active area, or already coupled into a waveguide. Differences in these definitions can make superficially similar specifications non-equivalent.
Quantum efficiency describes the number of collected charge carriers per incident photon, with internal and external definitions distinguishing absorbed photons from all incident photons. For an ordinary photodiode with no internal multiplication, external quantum efficiency gives
Expressing wavelength in micrometres gives approximately . Thus a hypothetical unity-efficiency detector at has a responsivity near . A value above does not by itself demonstrate internal gain; the charge delivered per unit optical energy also depends on photon energy.
Speed is limited by several mechanisms. Carriers require time to drift across the depleted absorber, and carriers generated outside it may arrive later by diffusion. Junction capacitance, package capacitance, and circuit impedance impose an electrical time constant. For a simple single-pole resistance–capacitance response,
The familiar relation , connecting 10–90 percent rise time to bandwidth, applies to an ideal single-pole response and is only an approximation for a complete detector.
Active area introduces a major trade-off. A large area captures a broad or poorly positioned optical beam more easily, but generally increases capacitance and often total leakage. Small-area detectors can reach much greater bandwidth, provided that the optical system concentrates light efficiently onto them. Waveguide integration offers another solution by confining the optical mode near a small electrical junction.
Linearity describes how closely photocurrent remains proportional to optical power. At high illumination, series resistance, heating, carrier screening of the internal field, or limitations of the readout amplifier can cause compression. At low illumination, dark-current uncertainty and noise limit the useful measurement range. Saturation power and dynamic range therefore belong to a detector-plus-circuit system, not simply to the semiconductor material.
Dark current, noise, and detection limits
A photodiode produces electrical fluctuations even when the desired optical signal is absent. Dark current is the current measured without incident signal light under specified operating conditions. It can arise from carrier generation in the depletion region, diffusion from neutral material, surface leakage, defects, and field-assisted tunneling. Different mechanisms dominate in different materials and voltage ranges. Temperature generally has a strong effect, but no single temperature-scaling rule applies to every photodiode.
The discrete arrival of charge produces shot noise. For an ideal unmultiplied photodiode carrying an average signal-plus-dark current , the current-noise variance within an equivalent noise bandwidth is
This relation assumes the relevant carrier arrivals can be treated as independent. Optical background contributes photocurrent and therefore noise even when it is not part of the desired signal. Once background shot noise dominates, reducing detector leakage may have little effect on overall sensitivity.
Resistive elements generate thermal noise. For a resistance represented as a parallel current-noise source at temperature ,
Readout amplifiers add current noise, voltage noise, and sometimes low-frequency excess noise. Junction capacitance converts amplifier voltage fluctuations into a frequency-dependent current contribution, making capacitance especially consequential in wideband systems. A detector with very low dark current can still perform poorly if connected to an unsuitable amplifier.
A common sensitivity metric is noise-equivalent power, or NEP. Expressed as a spectral quantity, it is the incident optical power per square root bandwidth that produces a signal equal to the detector system’s noise:
where is input-referred current-noise spectral density in . Lower NEP indicates greater sensitivity under the stated conditions. Specific detectivity, , normalizes sensitivity by detector area under specified conventions. Comparisons require attention to wavelength, modulation frequency, temperature, and which noise terms are included.
Avalanche gain changes this balance. An APD can amplify the optical signal before it encounters downstream amplifier noise, but multiplication adds excess noise and also amplifies some dark-current components. SPADs instead require a counting description: dark events, background counts, afterpulses, dead time, and timing uncertainty determine their detection performance.
Readout circuits and signal processing
A photodiode is fundamentally a current-generating sensor, so its practical behavior depends heavily on the circuit used to measure that current. The simplest readout places a resistor across the detector and measures the resulting voltage. Increasing resistance raises the signal voltage for a given photocurrent, but it also increases the electrical time constant associated with detector capacitance. At appreciable illumination, the changing diode voltage can further alter collection behavior or introduce nonlinearity.
A transimpedance amplifier converts input current into output voltage using feedback. In a common arrangement, an operational amplifier holds the photodiode terminal close to a reference voltage while the photocurrent flows through a feedback resistor . Within the amplifier’s useful operating range,
This allows substantial voltage gain without forcing the photodiode itself to undergo a comparably large voltage swing. A bias network can set the junction near zero voltage or at a selected reverse bias.
Detector capacitance, amplifier input capacitance, and finite amplifier gain complicate feedback stability. A capacitor placed across the feedback resistor is often used to control peaking and oscillation, at the cost of bandwidth. The achievable response depends on the complete feedback network rather than only on the nominal resistance–capacitance product of the photodiode. High-impedance measurements are also sensitive to board contamination, insulation leakage, humidity, and parasitic coupling.
Imaging and low-light systems often integrate charge rather than measure instantaneous current. A photocurrent charges or discharges a capacitor during a controlled exposure interval, giving approximately
where is integration time. Longer integration collects more signal but increases dark charge and risks saturation. Reset operations introduce their own noise, while correlated sampling can suppress certain reset and offset contributions. Conversion into digital values then introduces quantization and readout constraints.
Balanced detection uses two photodiodes and subtracts their outputs. When the optical signals and electrical paths are properly matched, common fluctuations can cancel while the desired differential signal remains. This technique is important in interferometry and coherent optical reception. It does not eliminate independent shot noise from the two detectors.
Some measurement systems modulate the light source and recover only the corresponding electrical component through synchronous detection. Others count individual SPAD events or measure their arrival times. These architectures extract different information from the same basic conversion of optical excitation into electrical charge.
Fabrication, packaging, and imaging arrays
Photodiode fabrication combines optical requirements with semiconductor process control. Junctions can be formed through diffusion, ion implantation, or epitaxial growth, while photolithography defines contacts, isolation regions, and active areas. Dopant concentration and layer thickness determine electric-field distribution, depletion depth, and breakdown behavior. Guard rings and related edge structures reduce electric-field crowding around the junction perimeter, where premature breakdown could otherwise occur.
Surface engineering is especially important because an exposed or poorly passivated surface can both lose photogenerated carriers and introduce leakage. Passivation layers reduce electrically active interface states, while antireflection coatings improve transmission into the absorber. Metal contacts must provide reliable electrical connection without obscuring too much of the optical aperture. Back illumination can avoid some front-side obstruction, but may require substrate thinning, additional processing, and careful mechanical support.
Packaging defines how radiation reaches the active material. A package window that is transparent in the visible may absorb ultraviolet or infrared light, so window composition must match the intended spectrum. Fiber-coupled detectors require alignment between the optical mode and the sensitive region. High-speed packages must minimize lead inductance and stray capacitance, whereas very low-current packages emphasize electrical insulation and protection against contamination. Cooling arrangements may be integrated when thermal noise or dark current is critical.
Arrays distribute photodetection across many elements to record spatial information. In a charge-coupled device, photocharge is accumulated in semiconductor potential wells and transferred through the device for readout; the photosensitive architecture need not be a simple discrete junction diode at each site. A CMOS image sensor commonly combines photodiodes with nearby transistors for reset, charge transfer, buffering, and selection. Each pixel therefore includes both a photosensitive element and circuitry or shared readout resources.
The pinned photodiode is particularly important in modern CMOS imaging. A suitably engineered surface layer helps control surface potential, while a buried region stores photocharge. A transfer gate moves that charge to a sensing node. This structure supports low dark current and low-noise readout when charge transfer and sampling are properly implemented.
Array design introduces additional constraints absent from a single detector: fill factor, pixel-to-pixel variation, optical and electrical crosstalk, full-well capacity, and readout throughput. Microlenses redirect light toward sensitive regions, while backside illumination increases the optical access available around front-side wiring. Defects or process variations can produce hot pixels and nonuniform response that require characterization and calibration.
Historical development
Photodiodes developed from the wider investigation of how light changes electrical behavior in materials. In 1839, Edmond Becquerel observed a photovoltaic effect in an electrochemical arrangement. In 1873, Willoughby Smith reported the light-dependent conductivity of selenium. These discoveries did not constitute modern junction photodiodes, but they established distinct ways in which radiation could generate electrical output or change a material’s conductivity.
Early light detectors included selenium cells and vacuum photoelectric tubes. Their operating principles differed: photoconductive cells changed resistance, whereas vacuum devices collected electrons emitted from an illuminated surface. The development of quantum mechanics supplied an explanation of photon absorption and electronic transitions, but practical solid-state photodiodes also required control over crystal purity, doping, contacts, and interfaces.
An important step toward junction-based devices occurred at Bell Laboratories in 1940, when Russell Ohl observed the light-sensitive behavior of a silicon structure containing a naturally formed p–n junction. The subsequent expansion of silicon and germanium technology made deliberately engineered junction devices increasingly practical. The same improvements in materials processing that supported the transistor also enabled lower-leakage and more reproducible optical detectors.
In 1954, Daryl Chapin, Calvin Fuller, and Gerald Pearson demonstrated a practical silicon solar cell at Bell Laboratories. Although designed for power conversion rather than signal detection, this work illustrated the growing maturity of semiconductor photovoltaic junctions. Later advances in high-purity material and controlled depletion layers supported p–i–n detectors with improved speed and collection efficiency. Avalanche structures added internal gain for detecting weak, rapidly varying signals.
Optical-fiber communications created strong demand for fast detectors at wavelengths where glass fibers transmit efficiently. This drove work on germanium and compound-semiconductor photodiodes, particularly InGaAs-based devices for the and regions. Detector development became closely linked to optical coupling, receiver electronics, and the noise requirements of high-bit-rate communication.
Imaging followed a partly parallel path. Charge-coupled devices emerged from work by Willard Boyle and George Smith in 1969. The pinned-photodiode structure was reported by Nobukazu Teranishi and colleagues in 1982 and subsequently became important in both CCD and CMOS imaging. Later development brought large SPAD arrays, integrated timing circuits, silicon-photonics receivers, and increasingly elaborate combinations of optical sensing and electronic processing.
Applications in communication, measurement, and sensing
In optical communication, a photodiode converts received optical signals into electrical signals that can be amplified and decoded. Direct-detection receivers commonly use p–i–n photodiodes or APDs to recover changes in optical power. Coherent receivers mix the received field with light from a local oscillator and use multiple photodiodes, often in balanced pairs, to recover amplitude and phase information. Detector bandwidth, responsivity, linearity, and noise all affect the receiver’s performance.
Scientific and industrial instruments use photodiodes for radiometry, spectroscopy, position measurement, and process monitoring. A calibrated detector can serve as an optical-power reference, but its spectral responsivity and spatial uniformity must be known. Spectrometers may direct different wavelengths onto different elements of a photodiode array. Position-sensitive detectors and quadrant photodiodes compare currents from separate regions to determine where a beam falls, enabling optical alignment, tracking, and displacement measurements.
Photodiodes also monitor light sources through feedback. A small fraction of a laser’s output can be directed onto a monitor diode, whose current is used to stabilize emitted power. Optical encoders detect patterned interruption or reflection of light to measure rotation or translation. Remote-control receivers combine a photodiode with amplification and filtering designed to identify a modulated infrared signal in the presence of ambient illumination.
In cameras, photodiode arrays provide the light-sensitive foundation of many electronic image sensors. Color filters or other wavelength-selective structures distinguish spectral bands, while integration and readout electronics convert collected charge into image data. These measurements subsequently support image processing and computer vision, but the photodiode itself responds to absorbed optical energy rather than recognizing objects or interpreting scenes.
Biomedical instruments use photodiodes to measure transmitted, reflected, or emitted light. Pulse oximeters, for example, compare signals obtained with red and infrared illumination as part of estimating arterial oxygen saturation. Fluorescence instruments detect light emitted by labeled samples, and optical analytical systems measure changes in absorption or scattering. The validity of such measurements depends on the complete instrument, calibration, and measurement conditions, not on detector sensitivity alone.
Time-resolved systems use fast photodiodes, APDs, or SPADs to measure optical pulse arrival. In lidar, travel time is related to distance by for a simple round trip through air when its refractive-index correction is neglected. Timing jitter, pulse shape, background light, and electronics all contribute to uncertainty. In particle and radiation instrumentation, photodiodes may detect either radiation directly or the light produced in a scintillator. Arrays of avalanche microcells, commonly called silicon photomultipliers, extend this approach to compact detectors with high internal gain.