Photovoltaic Mode

In photovoltaic mode, a photodiode operates without any external bias voltage, relying on the built-in potential difference across the p-n junction to separate photogenerated charge carriers. When photons with energy greater than the semiconductor bandgap are absorbed, they create electron-hole pairs primarily in or near the depletion region. These carriers are separated by the internal electric field: electrons drift toward the n-side and holes toward the p-side, generating a photocurrent that can produce a measurable voltage across the device terminals. This mode leverages the photovoltaic effect, similar to that in solar cells, but is tailored for light detection rather than efficient power conversion.[22]

The carrier dynamics in this mode involve both diffusion and drift processes. Generated electron-hole pairs in the neutral regions diffuse randomly until reaching the depletion region, where the strong built-in field sweeps them apart efficiently, minimizing recombination. Under short-circuit conditions (zero voltage across the device), the resulting current flows freely, while in open-circuit conditions, carrier accumulation builds up a voltage opposing further separation. The short-circuit current IscI_{sc}Isc​ is expressed as

where qqq is the elementary charge, η\etaη is the quantum efficiency, PPP is the incident optical power, AAA is the active area, and hνh \nuhν is the photon energy. The open-circuit voltage VocV_{oc}Voc​ is approximated by the diode equation

where kkk is Boltzmann's constant, TTT is the absolute temperature, and I0I_0I0​ is the dark saturation current.[22][12]

This operating mode offers distinct advantages, including very low noise due to negligible dark current and the ability to function in a self-powered manner without external circuitry. However, it has limitations such as slower response times compared to biased modes, as the absence of an external field reduces carrier collection efficiency and increases transit times. Photodiodes in photovoltaic mode are optimized for precise light detection with linear response to intensity, differing from solar cells which prioritize maximizing power output through larger areas and specific material choices.[23][12]

Photoconductive Mode

In photoconductive mode, a reverse bias voltage is applied across the photodiode, widening the depletion region compared to zero-bias operation and thereby improving the separation and collection efficiency of photogenerated electron-hole pairs while reducing junction capacitance.[24] This bias configuration causes the device to function as a light-dependent resistor, where the generated current varies directly with the intensity of incident light, enabling precise measurement of optical power.[25]

The photocurrent in this mode is expressed as Iph=R⋅PI_{ph} = R \cdot PIph​=R⋅P, where RRR is the responsivity (typically in A/W) and PPP is the incident optical power; the total current is then I=Iph+IdarkI = I_{ph} + I_{dark}I=Iph​+Idark​, with the dark current IdarkI_{dark}Idark​ increasing under the applied reverse bias voltage VrV_rVr​.[24]

The 3 dB bandwidth, which determines the frequency response, is influenced by the junction capacitance CjC_jCj​ and is commonly limited by the RC time constant, approximated as f3dB=12πRLCjf_{3dB} = \frac{1}{2\pi R_L C_j}f3dB​=2πRL​Cj​1​, where RLR_LRL​ is the load resistance; higher reverse bias reduces CjC_jCj​, extending the bandwidth for faster operation.[24]

Linearity in photoconductive mode is a key feature, with the output current maintaining a proportional relationship to light intensity across several orders of magnitude until saturation occurs, and the applied bias voltage enhances this by minimizing carrier recombination and diffusion effects that could introduce nonlinearity.[8]

This mode offers advantages such as superior speed and reduced capacitance relative to unbiased operation, rendering it ideal for high-frequency applications like fiber-optic communications and laser ranging systems.[26]

Semiconductor Materials

Silicon is the most widely used semiconductor material for photodiodes operating in the visible and near-infrared (NIR) spectrum, with a bandgap energy of 1.12 eV that enables efficient absorption of photons up to approximately 1100 nm.[27] Germanium, featuring a narrower bandgap of 0.67 eV, extends sensitivity into the infrared region up to about 1700 nm, making it suitable for mid-IR detection. Gallium arsenide (GaAs) photodiodes, with a bandgap of 1.43 eV, target NIR applications around 870 nm, offering higher electron mobility compared to silicon for faster response times.[25] Indium gallium arsenide (InGaAs), tunable with a bandgap around 0.75 eV, provides extended NIR coverage up to 1.7 μm, ideal for telecommunications wavelengths.[25]

Key material properties influencing photodiode performance include the wavelength-dependent absorption coefficient α(λ), which quantifies how strongly light is absorbed; the penetration depth, given by 1/α, determines the optimal placement of the p-n junction to maximize carrier collection efficiency.[27] For instance, silicon exhibits α values on the order of 10^4 cm⁻¹ at 800 nm, leading to shallow penetration depths of about 1 μm, while germanium's lower α in the IR necessitates thicker absorption layers.[28] Carrier mobility, measuring charge transport speed, and minority carrier lifetime, affecting recombination rates, directly impact response time; high-mobility materials like GaAs (electron mobility ~8500 cm²/V·s) enable bandwidths exceeding 10 GHz.[29]

Material selection for photodiodes hinges on the target wavelength range from ultraviolet to infrared, with silicon dominating UV-visible applications due to its broad absorption and temperature stability up to 150°C.[25] Temperature stability is critical, as bandgap energies decrease with rising temperature, shifting absorption edges; III-V compounds like InGaAs maintain performance better in harsh environments than germanium. Cost-performance trade-offs favor silicon for low-cost, high-volume visible detectors, while III-V materials such as GaAs and InGaAs are preferred for high-speed telecom despite higher fabrication expenses.[25]

Emerging materials as of 2025 include halide perovskites, which offer broadband absorption and high quantum efficiencies, often approaching 90% in optimized visible-NIR devices, though stability issues under humidity and heat limit commercial adoption.[30] Two-dimensional materials like graphene enable ultrafast detection with response times on the order of picoseconds to nanoseconds, benefiting from high carrier mobilities exceeding 10,000 cm²/V·s at room temperature, but challenges in bandgap engineering and integration persist.[21]

Doping levels in these materials form the p-n junction essential for carrier separation; in silicon, p-type doping typically uses boron at concentrations of 10^{15}-10^{18} cm^{-3} to create acceptor sites, while n-type doping employs phosphorus at similar levels to provide donor electrons.[31]

Fabrication Techniques

The fabrication of photodiodes begins with wafer preparation, where high-purity semiconductor substrates are produced to ensure minimal defects and uniform electrical properties. For silicon-based photodiodes, the Czochralski process is widely employed to grow single-crystal ingots from a molten silicon source, followed by slicing into wafers that serve as the foundation for device layers. Doping of these wafers to create n-type or p-type regions is achieved through diffusion, where dopant atoms like phosphorus or boron are introduced via thermal processes, or ion implantation, which accelerates dopant ions into the lattice for precise control over concentration profiles.[32] These steps are critical for establishing the base conductivity and junction characteristics essential to photodiode functionality.

Junction formation follows wafer preparation, particularly for PIN structures that require an intrinsic region to minimize capacitance and enhance speed. Epitaxial growth techniques such as metal-organic chemical vapor deposition (MOCVD) or molecular beam epitaxy (MBE) are used to deposit the intrinsic layer with atomic-level precision on the doped substrate, enabling low-defect interfaces and tailored bandgap properties. For instance, MOCVD facilitates uniform deposition of III-V materials like InGaAs for near-infrared detection, while MBE offers superior control for lattice-matched heterostructures in advanced devices.[33] These methods ensure the p-i-n junction's abrupt doping transitions, which are vital for efficient carrier collection.

Metallization and passivation steps protect the device and optimize optical coupling. Ohmic contacts, typically formed using titanium-aluminum (Ti/Al) stacks for n-type regions, are evaporated or sputtered onto the semiconductor surface followed by annealing to achieve low-resistance interfaces without rectification.[34] Passivation layers, such as silicon dioxide (SiO₂), are then deposited via plasma-enhanced chemical vapor deposition to encapsulate the junction and prevent surface recombination, while anti-reflective coatings like SiO₂ or multilayer stacks reduce reflection losses at the light-entry surface, improving quantum efficiency by up to 20-30% in silicon photodiodes.[35]

Packaging completes the fabrication by ensuring environmental stability and optical performance. Hermetic sealing, often in TO-can or ceramic packages, uses metal lids welded or soldered to the base to exclude moisture and contaminants, thereby enhancing long-term reliability in harsh conditions.[36] Integration with optics, such as aspheric lenses or windows, is incorporated during assembly to focus incident light onto the active area, minimizing divergence and maximizing coupling efficiency for applications like fiber-optic receivers.[37]

Yield considerations in photodiode fabrication emphasize defect density control to achieve high uniformity across wafers. Techniques like gettering during epitaxial growth remove impurities, targeting defect densities below 10⁸ cm⁻², while CMOS-compatible processes enable scaling to large arrays by leveraging standard lithography and backend steps for monolithic integration. This compatibility supports fabrication yields exceeding 90% for focal plane arrays in imaging systems.

PIN Photodiode

The PIN photodiode features a layered structure consisting of a p-type semiconductor region, an intrinsic (undoped or lightly doped) semiconductor region, and an n-type semiconductor region, denoted as p-i-n.[40] The intrinsic region, typically several micrometers thick, separates the heavily doped p and n layers and expands significantly under reverse bias to form a wide depletion zone.[12] This design contrasts with standard p-n photodiodes by minimizing carrier diffusion and enhancing the electric field uniformity across the absorption area.[22]

A primary advantage of the PIN structure is the reduced junction capacitance due to the extended depletion width, which lowers the RC time constant and improves high-frequency performance. The capacitance is approximated by the formula

where ϵ\epsilonϵ is the permittivity of the semiconductor material, AAA is the active area, WiW_iWi​ is the intrinsic region width, and WdW_dWd​ represents the depletion widths in the p and n regions (often negligible in heavily doped layers).[7] The wider depletion region also enables higher quantum efficiency across a broad spectral range from ultraviolet to infrared wavelengths, as more photogenerated carriers are collected before recombination.[4] Compared to p-n diodes, PIN photodiodes exhibit lower noise levels, primarily from reduced thermal noise associated with the lower capacitance and minimized dark current.[12]

In operation, particularly in photoconductive mode under reverse bias, the strong electric field in the intrinsic region sweeps photogenerated electron-hole pairs toward the respective contacts with minimal recombination, enabling efficient carrier collection.[11] This configuration supports high-speed applications, with typical bandwidths ranging from 10 to 100 GHz depending on the intrinsic layer thickness and material.[41] PIN photodiodes are briefly referenced here for their enhanced performance in reverse-biased photoconductive operation, as detailed in broader mode discussions.

Commonly employed in telecommunications receivers for optical signal detection, PIN photodiodes benefit from their balance of speed and sensitivity in fiber-optic systems.[22] However, they require higher reverse bias voltages—often tens of volts—to fully deplete the intrinsic region and achieve optimal performance, which can increase power consumption and circuit complexity.[42] Additionally, the multi-layer structure introduces greater fabrication challenges, including precise control of the intrinsic region's doping and thickness uniformity during epitaxial growth or diffusion processes.[43]

Avalanche Photodiode

Avalanche photodiodes (APDs) are specialized photodiodes that achieve internal current gain through carrier multiplication, enabling enhanced sensitivity for low-light detection in optical systems. Unlike standard photodiodes, APDs operate under high reverse bias to trigger impact ionization, amplifying the photocurrent while introducing specific noise characteristics. This gain mechanism makes APDs particularly valuable for applications requiring high signal-to-noise ratios, such as fiber-optic communications and photon counting.[19][44]

The structure of an APD typically incorporates a high-field multiplication region within a p-i-n configuration or utilizes separate absorption and multiplication layers to separate photon absorption from carrier multiplication, optimizing quantum efficiency and reducing noise. In the p-i-n based design, the intrinsic region is divided such that photogeneration occurs in a lower-field absorption zone, while multiplication happens in a narrower, high-field avalanche region under reverse bias exceeding 100 V. Separate absorption-multiplication structures, often denoted as SAM or SACM (separate absorption, charge, and multiplication), further enhance performance by tailoring material properties for specific wavelengths, such as InGaAs absorption layers paired with InP multiplication regions for near-infrared detection.[44][4][19]

The gain in APDs arises from impact ionization, where photogenerated carriers gain sufficient kinetic energy in the high electric field to ionize additional atoms, creating secondary electron-hole pairs that further multiply. This process yields a multiplication gain M=IoutIphM = \frac{I_\text{out}}{I_\text{ph}}M=Iph​Iout​​, where IoutI_\text{out}Iout​ is the output current and IphI_\text{ph}Iph​ is the primary photocurrent, with typical values ranging from 100 to 1000 depending on bias and material. The total output current can be expressed as

where IdarkI_\text{dark}Idark​ accounts for thermally generated carriers. However, the stochastic nature of ionization leads to excess noise, quantified by the noise factor F(M)≈MxF(M) \approx M^xF(M)≈Mx, with xxx as the excess noise index (typically 0.2–0.8 for optimized designs), which degrades the signal-to-noise ratio at high gains.[45][46]

APDs are classified by the initiating carrier and multiplication dynamics: electron-initiated types, common in InP-based devices, leverage higher electron ionization coefficients for lower noise, while hole-initiated variants in silicon exploit hole multiplication for visible-light applications. Reach-through APDs extend the depletion region to fully deplete the absorption layer, ensuring uniform field penetration and higher efficiency, whereas electron-hole APDs allow both carriers to contribute to multiplication, though this often increases noise due to mixed ionization rates. Silicon APDs favor electron initiation through doping profiles that prioritize electron avalanches, achieving gains up to 1000 with moderate noise.[47]

Despite their advantages, APDs face limitations from excess noise, which scales with gain and limits usable MMM to avoid signal degradation, as well as the risk of premature breakdown from field nonuniformities or defects. High operating voltages also necessitate precise bias control and often thermoelectric cooling to suppress thermal generation of dark current and maintain gain stability, particularly in arrays or high-temperature environments.[46][44]

Advances through 2025 have focused on low-noise superlattice APDs, incorporating type-II superlattices like InGaAs/GaAsSb for absorption and AlGaAsSb for multiplication, achieving gains over 100 with excess noise factors below 2 and gain-quantum efficiency products exceeding 3500% at 2 μm wavelengths, ideal for quantum sensing in mid-infrared regimes.[48][49] In 2025, further advancements include digital alloy AlAsSb/GaAsSb APDs demonstrating low dark current and noise for optical communications, and thin absorber AlInAsSb SACM APDs with suppressed dark currents at 2 μm.[50][51] These structures mitigate noise via engineered band alignments that favor single-carrier multiplication, enabling single-photon-level detection with reduced cooling requirements.

Responsivity and Sensitivity

Responsivity is a fundamental performance metric for photodiodes, defined as the ratio of the generated photocurrent IphI_{ph}Iph​ to the incident optical power PPP, expressed as R(λ)=IphPR(\lambda) = \frac{I_{ph}}{P}R(λ)=PIph​​.[52] This yields units of amperes per watt (A/W), quantifying the device's efficiency in converting light to electrical signal. The responsivity is wavelength-dependent, R(λ)R(\lambda)R(λ), and follows the relation R(λ)=qληhcR(\lambda) = \frac{q \lambda \eta}{h c}R(λ)=hcqλη​, where qqq is the elementary charge, λ\lambdaλ is the wavelength, η\etaη is the quantum efficiency, hhh is Planck's constant, and ccc is the speed of light; this equation highlights the linear scaling with photon energy and efficiency.[52] Spectral response curves, plotting R(λ)R(\lambda)R(λ) versus λ\lambdaλ, typically peak near the material's bandgap and drop sharply beyond the cutoff wavelength, such as for silicon photodiodes where maximum responsivity occurs around 900 nm.[52]

Quantum efficiency η(λ)\eta(\lambda)η(λ) is a key factor in responsivity, representing the ratio of charge carriers collected to incident photons, often reaching 70-90% in optimized devices.[53] External quantum efficiency accounts for losses like surface reflection, while internal quantum efficiency excludes these, focusing on absorption and collection within the active region. Material selection influences η(λ)\eta(\lambda)η(λ), with silicon offering high values in the visible to near-infrared due to its 1.12 eV bandgap, though broader spectra require materials like InGaAs for extended response.[52]

Sensitivity metrics extend beyond responsivity to characterize minimum detectable signals. The noise equivalent power (NEP) is the incident power yielding a signal-to-noise ratio of 1 in a 1 Hz bandwidth, typically in W/√Hz.[54] Specific detectivity D∗D^D∗, a normalized figure of merit, is given by D∗=AΔfNEPD^ = \frac{\sqrt{A \Delta f}}{NEP}D∗=NEPAΔf​​, with units cm √Hz/W, where AAA is the active area and Δf\Delta fΔf is the bandwidth; higher D∗D^*D∗ indicates superior low-light performance, often exceeding 10^{12} cm √Hz/W for silicon photodiodes.[54]

Responsivity and sensitivity are measured using calibrated monochromatic light sources, such as lasers or tunable lamps, to illuminate the device while monitoring photocurrent with a transimpedance amplifier under controlled bias.[24] Temperature affects these metrics, with responsivity varying with wavelength, typically decreasing by about 0.2-0.4% per °C in the visible and near-infrared regions due to bandgap widening, though it can be positive in the infrared.[55]

Optimization strategies enhance performance, including anti-reflective (AR) coatings that minimize Fresnel losses to boost external η\etaη up to 90% across target wavelengths.[57] Tailoring the absorption layer thickness and doping profiles further improves internal efficiency, while selecting materials matched to application wavelengths ensures peak responsivity, as in silicon for telecom bands near 850 nm.[58]

Speed and Noise Considerations

The speed of a photodiode is fundamentally limited by two primary factors: the RC time constant associated with the device's capacitance and load resistance, and the carrier transit time across the depletion region. The rise time τr\tau_rτr​ is approximated by τr=2.2RLCtotal\tau_r = 2.2 R_L C_{total}τr​=2.2RL​Ctotal​, where RLR_LRL​ is the load resistance and CtotalC_{total}Ctotal​ includes the junction capacitance and any parasitic capacitances, determining the temporal response in low-frequency applications. For high-speed operation, the bandwidth is further constrained by the carrier transit time ttr=W/vst_{tr} = W / v_sttr​=W/vs​, where WWW is the depletion layer width and vsv_svs​ is the saturation velocity of charge carriers (typically around 10710^7107 cm/s in silicon), as carriers must traverse the absorption region before collection.[59] These limits often result in 3 dB bandwidths ranging from GHz for optimized PIN structures to lower values in larger-area devices, with the junction capacitance scaling with active area and reverse bias reducing it by widening the depletion region.[25]

Noise in photodiodes arises from multiple sources that degrade the signal-to-noise ratio (SNR), particularly in low-light or high-speed scenarios. Shot noise, originating from the discrete nature of photocurrent and dark current, has a root-mean-square (rms) value given by ishot=2qIΔfi_{shot} = \sqrt{2 q I \Delta f}ishot​=2qIΔf​, where qqq is the electron charge, III is the total current, and Δf\Delta fΔf is the bandwidth; this Poisson-limited noise dominates under illumination and sets the fundamental quantum limit for detection.[60] Thermal (Johnson) noise, due to random thermal motion of charge carriers, contributes an rms current of 4kTΔf/R\sqrt{4 k T \Delta f / R}4kTΔf/R​, with kkk as Boltzmann's constant, TTT the temperature, and RRR the shunt or load resistance, becoming prominent in high-impedance circuits.[61] Additionally, 1/f (flicker) noise, which follows a 1/fα1/f^\alpha1/fα power spectral density (α≈1\alpha \approx 1α≈1) at low frequencies, stems from surface traps and material defects in semiconductors like InGaAs, significantly impacting baseband signals below 1 kHz.[62]

A key figure of merit for evaluating photodiode performance under noise constraints is the noise-equivalent bandwidth, which quantifies the effective frequency range where the integrated noise equals the shot noise in a 1 Hz band, guiding trade-offs such as increasing reverse bias to enhance speed (by reducing capacitance) at the cost of higher dark current and thus amplified shot noise.[25] For instance, higher bias voltages can achieve bandwidths exceeding 10 GHz but elevate thermal and shot noise densities, necessitating careful circuit design to maintain SNR above 20 dB in communication systems.[63] Mitigation strategies include reducing the active area to minimize junction capacitance (lowering RC limits and thermal noise) and employing transimpedance amplifiers (TIAs), which convert photocurrent to voltage with low-noise op-amps, achieving noise floors as low as 10 fA/√Hz while preserving bandwidth up to several GHz.[64][65]

Desired Photodiode Effects

In photodiodes, photogeneration occurs when incident photons with energy greater than the semiconductor bandgap are absorbed, primarily in the depletion region, creating electron-hole pairs or excitons that contribute to the photocurrent. This process is most efficient in the intrinsic or lightly doped regions where the built-in electric field separates the carriers before recombination, enabling high quantum efficiency in reverse-biased operation.[67]

Carrier collection in photodiodes relies on both drift and diffusion mechanisms, where the strong electric field in the depletion region accelerates minority carriers toward the contacts via drift, while diffusion aids in collecting carriers generated near the edges of the region. This field-assisted transport minimizes recombination losses, achieving quantum yields approaching unity and supporting fast response times in well-designed structures.[67]

Wavelength selectivity in photodiodes arises from the material's bandgap, which determines the cutoff wavelength beyond which absorption is negligible, allowing tailored sensitivity to specific spectral ranges.[68] The relationship is given by the equation

where EgE_gEg​ is the bandgap energy, hhh is Planck's constant, ccc is the speed of light, and λcutoff\lambda_{\text{cutoff}}λcutoff​ is the cutoff wavelength; for example, silicon photodiodes exhibit a sharp response drop around 1100 nm corresponding to Eg≈1.12E_g \approx 1.12Eg​≈1.12 eV.[68] This intrinsic filtering enhances applications requiring discrimination against longer wavelengths, such as in visible-light sensing.

Non-avalanche photoconductive gain in certain photodiodes, particularly those with trap states, results from carrier trapping that prolongs the lifetime of one carrier type, allowing the other to recirculate and amplify the photocurrent beyond unity gain. In wide-bandgap semiconductors like β-Ga₂O₃ used in Schottky photodiodes, this trapping by self-trapped holes enables gains exceeding 50 while maintaining reasonable linearity, beneficial for low-light detection without high-voltage operation.[69]

Thermal effects can positively influence photodiode stability through temperature-compensated bias designs, where controlled heating or circuit adjustments counteract variations in dark current and responsivity, ensuring consistent performance across operating ranges.

Unwanted Photodiode Effects

One significant unwanted effect in photodiodes is dark current, which refers to the small electric current that flows through the device in the complete absence of incident light. This current arises primarily from thermal generation of electron-hole pairs within the semiconductor material, particularly in the depletion region and adjacent quasi-neutral regions. The magnitude of the dark current can be approximated by the formula Idark=qni2A(Lpτp+Lnτn)/NdI_{\text{dark}} = q n_i^2 A \left( \frac{L_p}{\tau_p} + \frac{L_n}{\tau_n} \right) / N_dIdark​=qni2​A(τp​Lp​​+τn​Ln​​)/Nd​, where qqq is the elementary charge, nin_ini​ is the intrinsic carrier concentration, AAA is the junction area, LpL_pLp​ and LnL_nLn​ are the diffusion lengths for holes and electrons, τp\tau_pτp​ and τn\tau_nτn​ are the corresponding minority carrier lifetimes, and NdN_dNd​ is the donor concentration on the n-side (assuming an asymmetric junction).[70] Dark current exhibits strong temperature dependence, typically doubling for every 10°C increase in temperature due to the exponential rise in intrinsic carrier concentration with thermal energy.[23]

At high light intensities, photodiodes can experience saturation, where the output photocurrent no longer increases linearly with input optical power. This occurs due to gain compression mechanisms, such as space charge effects that slow carrier transport, and increased recombination losses in the active region, which reduce the collection efficiency of photogenerated carriers. In photodiode arrays, crosstalk represents another adverse effect, manifesting as unintended signal interference between adjacent elements. Optical crosstalk arises from light scattering or diffraction into neighboring pixels, while electrical crosstalk stems from capacitive coupling or substrate currents that propagate signals laterally through the shared semiconductor structure.[71]

Photodiodes are also susceptible to aging and long-term degradation, which can compromise performance over time. Exposure to radiation, such as in space environments, induces defects in the crystal lattice that increase dark current and reduce responsivity; for instance, silicon-based devices show significant degradation after proton irradiation doses equivalent to years in low-Earth orbit.[72] Humidity accelerates degradation by promoting moisture ingress, leading to corrosion at contacts or passivation layers and elevated leakage currents. These factors contribute to a reduced mean time between failures (MTBF), often estimated using accelerated life testing models that account for environmental stressors to predict operational reliability.

To mitigate these unwanted effects, several strategies are employed. Cooling the photodiode, such as through thermoelectric coolers, suppresses thermal generation and thus lowers dark current by factors of 10 or more per 30-50°C reduction.[73] Guard rings—doped regions surrounding the active junction—help alleviate edge effects by distributing the electric field more uniformly, preventing premature breakdown and reducing surface-related leakage currents.[74] For space applications, where radiation hardness is critical, gallium nitride (GaN)-based photodiodes offer superior tolerance; GaN's wide bandgap and defect-resistant structure provide better performance compared to silicon counterparts.[75]

Optical Sensing and Communication

Photodiodes play a pivotal role in optical sensing and communication systems, where they convert incident light into electrical signals for high-fidelity data transmission and environmental monitoring. In fiber optic receivers, indium gallium arsenide (InGaAs) PIN and avalanche photodiodes (APDs) are widely employed due to their sensitivity in the near-infrared range (900–1700 nm), enabling detection of modulated optical signals in telecommunications infrastructure. These devices support data rates from 10 Gbps to 400 Gbps in Ethernet applications, such as data center interconnects and long-haul networks, by providing low-noise amplification and high bandwidth exceeding 30 GHz. For instance, InGaAs APDs achieve a gain-bandwidth product of over 400 GHz, facilitating error-free operation in dense wavelength-division multiplexing (DWDM) systems. Bit error rates (BER) below 10^{-12} are routinely maintained through forward error correction (FEC) techniques, which correct pre-FEC BERs up to 10^{-3} by adding parity bits, ensuring reliable performance over distances exceeding 100 km without regeneration.

In visible-range applications like barcode scanners and light detection and ranging (LIDAR) systems, silicon photodiodes dominate owing to their cost-effectiveness, fast response times (rise times under 1 ns), and broad sensitivity from 400 nm to 1100 nm. Barcode scanners utilize silicon PIN photodiodes to detect reflected laser pulses from bar patterns, converting intensity variations into electrical signals for decoding at speeds up to thousands of scans per second. In LIDAR, silicon APDs or PIN photodiodes enable pulse detection by measuring the time-of-flight of short laser pulses (typically 10–100 ns duration), achieving sub-millimeter resolution in automotive and surveying applications through precise timing of photon arrival. These photodiodes offer quantum efficiencies above 80% in the visible spectrum, minimizing crosstalk and enabling compact, low-power designs.

Environmental sensing leverages specialized photodiodes for non-contact monitoring of atmospheric conditions. Gallium nitride (GaN) photodiodes excel in ultraviolet (UV) monitors, exhibiting solar-blind operation (cutoff below 365 nm) and high responsivity (up to 0.2 A/W at 254 nm), making them ideal for detecting ozone levels or solar UV index without interference from visible light. These devices operate in photovoltaic mode for low-power, real-time measurement in outdoor stations, with detectivities exceeding 10^{12} Jones for trace UV flux monitoring. For gas detection, photodiodes integrated into absorption spectroscopy systems quantify species like methane or CO2 by measuring light attenuation at specific wavelengths; silicon or InGaAs photodiodes paired with tunable diode lasers achieve parts-per-billion sensitivity through differential detection, reducing noise from ambient light.

Integration of photodiodes with vertical-cavity surface-emitting lasers (VCSELs) in optical transceivers enhances efficiency for short-reach links in 5G and emerging 6G networks. InGaAs or silicon photodiodes are co-packaged with 850 nm or 1310 nm VCSELs in pluggable modules (e.g., QSFP-DD), supporting bidirectional data rates up to 400 Gbps over multimode fiber with power consumption below 10 W per channel. In 5G fronthaul, these transceivers transport digitized radio signals between baseband units and remote radio heads, achieving latencies under 100 μs via time-division duplexing. For 6G prototypes, hybrid integration on silicon photonics platforms incorporates PIN photodiodes for analog fronthaul, enabling terabit-per-second aggregate capacity while meeting power-over-fiber requirements for remote units.

A notable case study is the deployment of photodiodes in submarine fiber optic cables, which underpin over 99% of global internet traffic. At cable landing stations, high-sensitivity InGaAs APDs serve as receivers in terminal equipment, demodulating wavelength-multiplexed signals at rates up to 400 Gbps per channel across transoceanic spans exceeding 10,000 km. While modern repeaters primarily use erbium-doped fiber amplifiers (EDFAs) for all-optical regeneration every 50–80 km, legacy systems and monitoring nodes incorporate PIN/APD pairs for fault detection and performance verification, ensuring BER below 10^{-12} with FEC to support seamless intercontinental connectivity. This infrastructure, exemplified by systems like the MAREA cable (Microsoft/Facebook, 2018), relies on photodiode arrays for multiplexing up to 256 channels, facilitating petabit-scale throughput for cloud services and real-time applications.

Imaging and Scientific Instrumentation

In imaging applications, complementary metal-oxide-semiconductor (CMOS) image sensors employ active pixel sensors (APS) that integrate photodiodes as the primary light-detecting elements, enabling compact and cost-effective designs for consumer cameras such as those in smartphones and digital cameras. Each pixel in a CMOS APS consists of a photodiode paired with one or more transistors for amplification and readout, allowing for on-chip signal processing that reduces the need for external circuitry and improves power efficiency compared to traditional charge-coupled devices (CCDs).[76][77] These sensors dominate consumer markets due to their ability to capture high-resolution images at video rates, with fill factors often exceeding 60% through microlens arrays that direct light onto the photodiode active area.[78]

In scientific instrumentation, photodiode arrays play a crucial role in spectroscopy, particularly in monochromators where silicon (Si) and germanium (Ge) arrays enable precise material analysis across visible and infrared spectra. Si photodiode arrays detect wavelengths from ultraviolet to near-infrared (approximately 200–1100 nm), while Ge variants extend sensitivity into the mid-infrared (up to 1800 nm), facilitating applications like Raman spectroscopy and environmental monitoring by dispersing light through a monochromator and measuring intensity at multiple channels simultaneously.[79][80] These arrays provide high spatial resolution and low crosstalk, essential for resolving spectral lines in chemical composition studies, with quantum efficiencies often reaching 80–90% in optimized Si configurations.[81]

Medical applications leverage photodiodes for non-invasive diagnostics, such as in pulse oximeters operating in transmission mode, where red and near-infrared light passes through tissue like a fingertip, and the transmitted intensity is detected by a photodiode to calculate blood oxygen saturation via the ratio of absorbed light at dual wavelengths (typically 660 nm and 940 nm).[82][83] In fluorescence microscopy, photodiodes, including avalanche variants, detect emitted light from fluorophores excited by lasers, enabling high-sensitivity imaging of cellular structures with sub-micron resolution; for instance, single-photon avalanche diodes (SPADs) integrated into arrays achieve picosecond timing for fluorescence lifetime measurements.[84][85]

In particle physics, silicon photomultipliers (SiPMs)—compact arrays of avalanche photodiodes—serve as key detectors in experiments at CERN, such as the CMS upgrade at the Large Hadron Collider, where they couple to scintillating fibers or crystals to timestamp particle interactions with sub-nanosecond precision and detect low-light yields from high-energy collisions.[86][87] SiPMs offer photon detection efficiencies up to 50%, magnetic field insensitivity, and robustness in radiation environments, outperforming traditional photomultiplier tubes in calorimetry and tracking systems.[88][89]

Photodiode Arrays

Photodiode arrays integrate multiple photodiodes into structured arrangements to enable parallel light detection across spatial dimensions, facilitating applications such as scanning and imaging. Linear one-dimensional (1D) arrays typically consist of 512 to 1024 silicon photodiodes, each approximately 25 μm wide and 2 mm high, arranged in a row for tasks like spectrophotometry where spectral dispersion is captured along a single axis.[94][95] Two-dimensional (2D) arrays expand this to grids with M rows and N columns, often using charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) architectures to form imaging planes in cameras, where horizontally oriented CCD lines or pixel matrices collect light for full-frame capture.[96][97]

Readout mechanisms in photodiode arrays vary by architecture to transfer accumulated charge efficiently. In CCD-based arrays, charge is serially shifted through the array via coupled gates, enabling low-noise transfer but requiring global clocking that can limit speed in large formats.[98] In contrast, CMOS active pixel sensor (APS) arrays incorporate amplifiers and addressing circuitry at each pixel, allowing random access readout for higher speed and integration of on-chip processing, though with potentially higher noise from transistor variability.[77][99] Passive-pixel sensors (PPS), an earlier CMOS variant, rely on shared column gates for readout, offering simpler fabrication but suffering from higher fixed-pattern noise compared to APS due to the lack of per-pixel amplification.[100]

Scalability of photodiode arrays has advanced to support resolutions exceeding 100 megapixels in CMOS implementations, driven by shrinking pixel pitches down to 2-4 μm while maintaining performance.[101] Fill factors, the ratio of active light-sensitive area to total pixel area, routinely surpass 90% through the integration of microlens arrays that focus incident light onto the photodiode, minimizing losses from surrounding circuitry.[102] This enhancement is particularly critical in APS designs where transistor overhead reduces intrinsic photosensitive area.

Key challenges in photodiode arrays include achieving uniformity across elements and preventing blooming. Pixel-to-pixel gain variations are typically held below 10% in CCD arrays through precise fabrication, but CMOS APS can exhibit higher fixed-pattern noise from threshold mismatches, necessitating calibration techniques like correlated double sampling.[103] Blooming, the overflow of excess charge into adjacent pixels during high illumination, is mitigated by anti-blooming gates or vertical overflow drains in CCD structures, limiting spillover to neighboring diodes and preserving image fidelity.[104][105]

Recent advances through 2025 have focused on stacked architectures and back-side illumination (BSI) to overcome traditional limitations. Three-wafer stacked CMOS sensors with BSI photodiodes enable global shutter operation by separating charge collection from readout circuitry, achieving shutter efficiencies over 99.9% without rolling artifacts in high-speed imaging.[106] BSI configurations illuminate the photodiode from the backside, bypassing front-side wiring shadows to boost quantum efficiency (QE) to above 90% in the visible spectrum, particularly for small pixels in megapixel arrays.[107] These innovations, including 2.2 μm pixel stacked BSI global shutter sensors, support compact, high-dynamic-range modules for demanding environments like machine vision.[108]

Pinned Photodiode

The pinned photodiode (PPD) is a buried photodiode structure essential to modern charge-coupled device (CCD) and complementary metal-oxide-semiconductor (CMOS) image sensors, featuring a p⁺/n/p configuration that pins the surface potential to minimize noise and defects.[109] The n-type region serves as the charge collection volume within a p-type epitaxial layer or substrate, while the shallow p⁺ pinning layer at the surface accumulates holes to prevent depletion and isolate the active area from interface traps at the silicon-oxide boundary.[110] This design enables efficient photoelectron accumulation and transfer, distinguishing it from simpler p-n junction photodiodes.[111]

Developed in the early 1980s by Nobukazu Teranishi and colleagues at NEC Corporation, the PPD was initially introduced for interline-transfer CCDs to suppress vertical smear and reduce dark current generation, marking a pivotal advancement over surface-channel photodiodes.[18] Its adaptation to CMOS active pixel sensors in the mid-1990s, led by Eric Fossum at NASA's Jet Propulsion Laboratory, revolutionized consumer and scientific imaging by integrating low-noise detection with on-chip amplification.[109] The structure's physics relies on the pinned potential maintaining a stable electric field, which confines photogenerated electrons in the n-well during exposure while allowing complete depletion for charge transfer.[112]

In typical operation within a four-transistor (4T) CMOS pixel, the PPD integrates photocharge under reverse bias from the substrate contact, with electrons collected in the n-region potential pocket as light absorption creates electron-hole pairs—holes drifting to the p-substrate and electrons to the n-buried layer.[111] Readout involves pulsing the n⁺-implanted transfer gate to form a channel under the gate oxide, rapidly transferring the charge packet to an adjacent floating diffusion node without residual lag, followed by source-follower buffering and correlated double sampling to subtract reset noise.[110] The pinning layer's hole accumulation suppresses trap-assisted generation, yielding dark current densities as low as 0.1–1 e⁻/pixel/s at room temperature in optimized processes.[112]

Key advantages include near-complete charge transfer efficiency (>99.99%), eliminating image lag seen in partially depleted diodes, and inherently low read noise (down to sub-electron levels with CDS), which supports high dynamic range exceeding 70 dB in back-illuminated implementations.[109] The shallow p⁺ layer enhances blue-light sensitivity by reducing absorption losses near the surface, while compatibility with standard CMOS fabrication enables cost-effective scaling to megapixel arrays.[110] Limitations involve potential pinning voltage variability affecting full well capacity (typically 10,000–50,000 electrons), addressed in advanced variants through epitaxial thickness optimization.[111]

Recent innovations extend PPDs to non-traditional substrates, such as thin-film polycrystalline silicon on glass, where a stacked p⁺/i/n/p structure preserves pinning benefits for flexible sensors, achieving conversion gains of 50–100 μV/e⁻ and dark currents below 10 fA/cm² while enabling full charge transfer for noise suppression.[113] Fully depleted PPDs, biased via substrate reverse voltage in high-resistivity silicon, further improve near-infrared quantum efficiency (>80%) and crosstalk reduction in time-of-flight applications.[114] These configurations underscore the PPD's enduring impact, powering over 90% of commercial image sensors as of 2020.[109]

Photovoltaic Mode

In photovoltaic mode, a photodiode operates without any external bias voltage, relying on the built-in potential difference across the p-n junction to separate photogenerated charge carriers. When photons with energy greater than the semiconductor bandgap are absorbed, they create electron-hole pairs primarily in or near the depletion region. These carriers are separated by the internal electric field: electrons drift toward the n-side and holes toward the p-side, generating a photocurrent that can produce a measurable voltage across the device terminals. This mode leverages the photovoltaic effect, similar to that in solar cells, but is tailored for light detection rather than efficient power conversion.[22]

The carrier dynamics in this mode involve both diffusion and drift processes. Generated electron-hole pairs in the neutral regions diffuse randomly until reaching the depletion region, where the strong built-in field sweeps them apart efficiently, minimizing recombination. Under short-circuit conditions (zero voltage across the device), the resulting current flows freely, while in open-circuit conditions, carrier accumulation builds up a voltage opposing further separation. The short-circuit current IscI_{sc}Isc​ is expressed as

where qqq is the elementary charge, η\etaη is the quantum efficiency, PPP is the incident optical power, AAA is the active area, and hνh \nuhν is the photon energy. The open-circuit voltage VocV_{oc}Voc​ is approximated by the diode equation

where kkk is Boltzmann's constant, TTT is the absolute temperature, and I0I_0I0​ is the dark saturation current.[22][12]

This operating mode offers distinct advantages, including very low noise due to negligible dark current and the ability to function in a self-powered manner without external circuitry. However, it has limitations such as slower response times compared to biased modes, as the absence of an external field reduces carrier collection efficiency and increases transit times. Photodiodes in photovoltaic mode are optimized for precise light detection with linear response to intensity, differing from solar cells which prioritize maximizing power output through larger areas and specific material choices.[23][12]

Photoconductive Mode

In photoconductive mode, a reverse bias voltage is applied across the photodiode, widening the depletion region compared to zero-bias operation and thereby improving the separation and collection efficiency of photogenerated electron-hole pairs while reducing junction capacitance.[24] This bias configuration causes the device to function as a light-dependent resistor, where the generated current varies directly with the intensity of incident light, enabling precise measurement of optical power.[25]

The photocurrent in this mode is expressed as Iph=R⋅PI_{ph} = R \cdot PIph​=R⋅P, where RRR is the responsivity (typically in A/W) and PPP is the incident optical power; the total current is then I=Iph+IdarkI = I_{ph} + I_{dark}I=Iph​+Idark​, with the dark current IdarkI_{dark}Idark​ increasing under the applied reverse bias voltage VrV_rVr​.[24]

The 3 dB bandwidth, which determines the frequency response, is influenced by the junction capacitance CjC_jCj​ and is commonly limited by the RC time constant, approximated as f3dB=12πRLCjf_{3dB} = \frac{1}{2\pi R_L C_j}f3dB​=2πRL​Cj​1​, where RLR_LRL​ is the load resistance; higher reverse bias reduces CjC_jCj​, extending the bandwidth for faster operation.[24]

Linearity in photoconductive mode is a key feature, with the output current maintaining a proportional relationship to light intensity across several orders of magnitude until saturation occurs, and the applied bias voltage enhances this by minimizing carrier recombination and diffusion effects that could introduce nonlinearity.[8]

This mode offers advantages such as superior speed and reduced capacitance relative to unbiased operation, rendering it ideal for high-frequency applications like fiber-optic communications and laser ranging systems.[26]

Semiconductor Materials

Silicon is the most widely used semiconductor material for photodiodes operating in the visible and near-infrared (NIR) spectrum, with a bandgap energy of 1.12 eV that enables efficient absorption of photons up to approximately 1100 nm.[27] Germanium, featuring a narrower bandgap of 0.67 eV, extends sensitivity into the infrared region up to about 1700 nm, making it suitable for mid-IR detection. Gallium arsenide (GaAs) photodiodes, with a bandgap of 1.43 eV, target NIR applications around 870 nm, offering higher electron mobility compared to silicon for faster response times.[25] Indium gallium arsenide (InGaAs), tunable with a bandgap around 0.75 eV, provides extended NIR coverage up to 1.7 μm, ideal for telecommunications wavelengths.[25]

Key material properties influencing photodiode performance include the wavelength-dependent absorption coefficient α(λ), which quantifies how strongly light is absorbed; the penetration depth, given by 1/α, determines the optimal placement of the p-n junction to maximize carrier collection efficiency.[27] For instance, silicon exhibits α values on the order of 10^4 cm⁻¹ at 800 nm, leading to shallow penetration depths of about 1 μm, while germanium's lower α in the IR necessitates thicker absorption layers.[28] Carrier mobility, measuring charge transport speed, and minority carrier lifetime, affecting recombination rates, directly impact response time; high-mobility materials like GaAs (electron mobility ~8500 cm²/V·s) enable bandwidths exceeding 10 GHz.[29]

Material selection for photodiodes hinges on the target wavelength range from ultraviolet to infrared, with silicon dominating UV-visible applications due to its broad absorption and temperature stability up to 150°C.[25] Temperature stability is critical, as bandgap energies decrease with rising temperature, shifting absorption edges; III-V compounds like InGaAs maintain performance better in harsh environments than germanium. Cost-performance trade-offs favor silicon for low-cost, high-volume visible detectors, while III-V materials such as GaAs and InGaAs are preferred for high-speed telecom despite higher fabrication expenses.[25]

Emerging materials as of 2025 include halide perovskites, which offer broadband absorption and high quantum efficiencies, often approaching 90% in optimized visible-NIR devices, though stability issues under humidity and heat limit commercial adoption.[30] Two-dimensional materials like graphene enable ultrafast detection with response times on the order of picoseconds to nanoseconds, benefiting from high carrier mobilities exceeding 10,000 cm²/V·s at room temperature, but challenges in bandgap engineering and integration persist.[21]

Doping levels in these materials form the p-n junction essential for carrier separation; in silicon, p-type doping typically uses boron at concentrations of 10^{15}-10^{18} cm^{-3} to create acceptor sites, while n-type doping employs phosphorus at similar levels to provide donor electrons.[31]

Fabrication Techniques

The fabrication of photodiodes begins with wafer preparation, where high-purity semiconductor substrates are produced to ensure minimal defects and uniform electrical properties. For silicon-based photodiodes, the Czochralski process is widely employed to grow single-crystal ingots from a molten silicon source, followed by slicing into wafers that serve as the foundation for device layers. Doping of these wafers to create n-type or p-type regions is achieved through diffusion, where dopant atoms like phosphorus or boron are introduced via thermal processes, or ion implantation, which accelerates dopant ions into the lattice for precise control over concentration profiles.[32] These steps are critical for establishing the base conductivity and junction characteristics essential to photodiode functionality.

Junction formation follows wafer preparation, particularly for PIN structures that require an intrinsic region to minimize capacitance and enhance speed. Epitaxial growth techniques such as metal-organic chemical vapor deposition (MOCVD) or molecular beam epitaxy (MBE) are used to deposit the intrinsic layer with atomic-level precision on the doped substrate, enabling low-defect interfaces and tailored bandgap properties. For instance, MOCVD facilitates uniform deposition of III-V materials like InGaAs for near-infrared detection, while MBE offers superior control for lattice-matched heterostructures in advanced devices.[33] These methods ensure the p-i-n junction's abrupt doping transitions, which are vital for efficient carrier collection.

Metallization and passivation steps protect the device and optimize optical coupling. Ohmic contacts, typically formed using titanium-aluminum (Ti/Al) stacks for n-type regions, are evaporated or sputtered onto the semiconductor surface followed by annealing to achieve low-resistance interfaces without rectification.[34] Passivation layers, such as silicon dioxide (SiO₂), are then deposited via plasma-enhanced chemical vapor deposition to encapsulate the junction and prevent surface recombination, while anti-reflective coatings like SiO₂ or multilayer stacks reduce reflection losses at the light-entry surface, improving quantum efficiency by up to 20-30% in silicon photodiodes.[35]

Packaging completes the fabrication by ensuring environmental stability and optical performance. Hermetic sealing, often in TO-can or ceramic packages, uses metal lids welded or soldered to the base to exclude moisture and contaminants, thereby enhancing long-term reliability in harsh conditions.[36] Integration with optics, such as aspheric lenses or windows, is incorporated during assembly to focus incident light onto the active area, minimizing divergence and maximizing coupling efficiency for applications like fiber-optic receivers.[37]

Yield considerations in photodiode fabrication emphasize defect density control to achieve high uniformity across wafers. Techniques like gettering during epitaxial growth remove impurities, targeting defect densities below 10⁸ cm⁻², while CMOS-compatible processes enable scaling to large arrays by leveraging standard lithography and backend steps for monolithic integration. This compatibility supports fabrication yields exceeding 90% for focal plane arrays in imaging systems.

PIN Photodiode

The PIN photodiode features a layered structure consisting of a p-type semiconductor region, an intrinsic (undoped or lightly doped) semiconductor region, and an n-type semiconductor region, denoted as p-i-n.[40] The intrinsic region, typically several micrometers thick, separates the heavily doped p and n layers and expands significantly under reverse bias to form a wide depletion zone.[12] This design contrasts with standard p-n photodiodes by minimizing carrier diffusion and enhancing the electric field uniformity across the absorption area.[22]

A primary advantage of the PIN structure is the reduced junction capacitance due to the extended depletion width, which lowers the RC time constant and improves high-frequency performance. The capacitance is approximated by the formula

where ϵ\epsilonϵ is the permittivity of the semiconductor material, AAA is the active area, WiW_iWi​ is the intrinsic region width, and WdW_dWd​ represents the depletion widths in the p and n regions (often negligible in heavily doped layers).[7] The wider depletion region also enables higher quantum efficiency across a broad spectral range from ultraviolet to infrared wavelengths, as more photogenerated carriers are collected before recombination.[4] Compared to p-n diodes, PIN photodiodes exhibit lower noise levels, primarily from reduced thermal noise associated with the lower capacitance and minimized dark current.[12]

In operation, particularly in photoconductive mode under reverse bias, the strong electric field in the intrinsic region sweeps photogenerated electron-hole pairs toward the respective contacts with minimal recombination, enabling efficient carrier collection.[11] This configuration supports high-speed applications, with typical bandwidths ranging from 10 to 100 GHz depending on the intrinsic layer thickness and material.[41] PIN photodiodes are briefly referenced here for their enhanced performance in reverse-biased photoconductive operation, as detailed in broader mode discussions.

Commonly employed in telecommunications receivers for optical signal detection, PIN photodiodes benefit from their balance of speed and sensitivity in fiber-optic systems.[22] However, they require higher reverse bias voltages—often tens of volts—to fully deplete the intrinsic region and achieve optimal performance, which can increase power consumption and circuit complexity.[42] Additionally, the multi-layer structure introduces greater fabrication challenges, including precise control of the intrinsic region's doping and thickness uniformity during epitaxial growth or diffusion processes.[43]

Avalanche Photodiode

Avalanche photodiodes (APDs) are specialized photodiodes that achieve internal current gain through carrier multiplication, enabling enhanced sensitivity for low-light detection in optical systems. Unlike standard photodiodes, APDs operate under high reverse bias to trigger impact ionization, amplifying the photocurrent while introducing specific noise characteristics. This gain mechanism makes APDs particularly valuable for applications requiring high signal-to-noise ratios, such as fiber-optic communications and photon counting.[19][44]

The structure of an APD typically incorporates a high-field multiplication region within a p-i-n configuration or utilizes separate absorption and multiplication layers to separate photon absorption from carrier multiplication, optimizing quantum efficiency and reducing noise. In the p-i-n based design, the intrinsic region is divided such that photogeneration occurs in a lower-field absorption zone, while multiplication happens in a narrower, high-field avalanche region under reverse bias exceeding 100 V. Separate absorption-multiplication structures, often denoted as SAM or SACM (separate absorption, charge, and multiplication), further enhance performance by tailoring material properties for specific wavelengths, such as InGaAs absorption layers paired with InP multiplication regions for near-infrared detection.[44][4][19]

The gain in APDs arises from impact ionization, where photogenerated carriers gain sufficient kinetic energy in the high electric field to ionize additional atoms, creating secondary electron-hole pairs that further multiply. This process yields a multiplication gain M=IoutIphM = \frac{I_\text{out}}{I_\text{ph}}M=Iph​Iout​​, where IoutI_\text{out}Iout​ is the output current and IphI_\text{ph}Iph​ is the primary photocurrent, with typical values ranging from 100 to 1000 depending on bias and material. The total output current can be expressed as

where IdarkI_\text{dark}Idark​ accounts for thermally generated carriers. However, the stochastic nature of ionization leads to excess noise, quantified by the noise factor F(M)≈MxF(M) \approx M^xF(M)≈Mx, with xxx as the excess noise index (typically 0.2–0.8 for optimized designs), which degrades the signal-to-noise ratio at high gains.[45][46]

APDs are classified by the initiating carrier and multiplication dynamics: electron-initiated types, common in InP-based devices, leverage higher electron ionization coefficients for lower noise, while hole-initiated variants in silicon exploit hole multiplication for visible-light applications. Reach-through APDs extend the depletion region to fully deplete the absorption layer, ensuring uniform field penetration and higher efficiency, whereas electron-hole APDs allow both carriers to contribute to multiplication, though this often increases noise due to mixed ionization rates. Silicon APDs favor electron initiation through doping profiles that prioritize electron avalanches, achieving gains up to 1000 with moderate noise.[47]

Despite their advantages, APDs face limitations from excess noise, which scales with gain and limits usable MMM to avoid signal degradation, as well as the risk of premature breakdown from field nonuniformities or defects. High operating voltages also necessitate precise bias control and often thermoelectric cooling to suppress thermal generation of dark current and maintain gain stability, particularly in arrays or high-temperature environments.[46][44]

Advances through 2025 have focused on low-noise superlattice APDs, incorporating type-II superlattices like InGaAs/GaAsSb for absorption and AlGaAsSb for multiplication, achieving gains over 100 with excess noise factors below 2 and gain-quantum efficiency products exceeding 3500% at 2 μm wavelengths, ideal for quantum sensing in mid-infrared regimes.[48][49] In 2025, further advancements include digital alloy AlAsSb/GaAsSb APDs demonstrating low dark current and noise for optical communications, and thin absorber AlInAsSb SACM APDs with suppressed dark currents at 2 μm.[50][51] These structures mitigate noise via engineered band alignments that favor single-carrier multiplication, enabling single-photon-level detection with reduced cooling requirements.

Responsivity and Sensitivity

Responsivity is a fundamental performance metric for photodiodes, defined as the ratio of the generated photocurrent IphI_{ph}Iph​ to the incident optical power PPP, expressed as R(λ)=IphPR(\lambda) = \frac{I_{ph}}{P}R(λ)=PIph​​.[52] This yields units of amperes per watt (A/W), quantifying the device's efficiency in converting light to electrical signal. The responsivity is wavelength-dependent, R(λ)R(\lambda)R(λ), and follows the relation R(λ)=qληhcR(\lambda) = \frac{q \lambda \eta}{h c}R(λ)=hcqλη​, where qqq is the elementary charge, λ\lambdaλ is the wavelength, η\etaη is the quantum efficiency, hhh is Planck's constant, and ccc is the speed of light; this equation highlights the linear scaling with photon energy and efficiency.[52] Spectral response curves, plotting R(λ)R(\lambda)R(λ) versus λ\lambdaλ, typically peak near the material's bandgap and drop sharply beyond the cutoff wavelength, such as for silicon photodiodes where maximum responsivity occurs around 900 nm.[52]

Quantum efficiency η(λ)\eta(\lambda)η(λ) is a key factor in responsivity, representing the ratio of charge carriers collected to incident photons, often reaching 70-90% in optimized devices.[53] External quantum efficiency accounts for losses like surface reflection, while internal quantum efficiency excludes these, focusing on absorption and collection within the active region. Material selection influences η(λ)\eta(\lambda)η(λ), with silicon offering high values in the visible to near-infrared due to its 1.12 eV bandgap, though broader spectra require materials like InGaAs for extended response.[52]

Sensitivity metrics extend beyond responsivity to characterize minimum detectable signals. The noise equivalent power (NEP) is the incident power yielding a signal-to-noise ratio of 1 in a 1 Hz bandwidth, typically in W/√Hz.[54] Specific detectivity D∗D^D∗, a normalized figure of merit, is given by D∗=AΔfNEPD^ = \frac{\sqrt{A \Delta f}}{NEP}D∗=NEPAΔf​​, with units cm √Hz/W, where AAA is the active area and Δf\Delta fΔf is the bandwidth; higher D∗D^*D∗ indicates superior low-light performance, often exceeding 10^{12} cm √Hz/W for silicon photodiodes.[54]

Responsivity and sensitivity are measured using calibrated monochromatic light sources, such as lasers or tunable lamps, to illuminate the device while monitoring photocurrent with a transimpedance amplifier under controlled bias.[24] Temperature affects these metrics, with responsivity varying with wavelength, typically decreasing by about 0.2-0.4% per °C in the visible and near-infrared regions due to bandgap widening, though it can be positive in the infrared.[55]

Optimization strategies enhance performance, including anti-reflective (AR) coatings that minimize Fresnel losses to boost external η\etaη up to 90% across target wavelengths.[57] Tailoring the absorption layer thickness and doping profiles further improves internal efficiency, while selecting materials matched to application wavelengths ensures peak responsivity, as in silicon for telecom bands near 850 nm.[58]

Speed and Noise Considerations

The speed of a photodiode is fundamentally limited by two primary factors: the RC time constant associated with the device's capacitance and load resistance, and the carrier transit time across the depletion region. The rise time τr\tau_rτr​ is approximated by τr=2.2RLCtotal\tau_r = 2.2 R_L C_{total}τr​=2.2RL​Ctotal​, where RLR_LRL​ is the load resistance and CtotalC_{total}Ctotal​ includes the junction capacitance and any parasitic capacitances, determining the temporal response in low-frequency applications. For high-speed operation, the bandwidth is further constrained by the carrier transit time ttr=W/vst_{tr} = W / v_sttr​=W/vs​, where WWW is the depletion layer width and vsv_svs​ is the saturation velocity of charge carriers (typically around 10710^7107 cm/s in silicon), as carriers must traverse the absorption region before collection.[59] These limits often result in 3 dB bandwidths ranging from GHz for optimized PIN structures to lower values in larger-area devices, with the junction capacitance scaling with active area and reverse bias reducing it by widening the depletion region.[25]

Noise in photodiodes arises from multiple sources that degrade the signal-to-noise ratio (SNR), particularly in low-light or high-speed scenarios. Shot noise, originating from the discrete nature of photocurrent and dark current, has a root-mean-square (rms) value given by ishot=2qIΔfi_{shot} = \sqrt{2 q I \Delta f}ishot​=2qIΔf​, where qqq is the electron charge, III is the total current, and Δf\Delta fΔf is the bandwidth; this Poisson-limited noise dominates under illumination and sets the fundamental quantum limit for detection.[60] Thermal (Johnson) noise, due to random thermal motion of charge carriers, contributes an rms current of 4kTΔf/R\sqrt{4 k T \Delta f / R}4kTΔf/R​, with kkk as Boltzmann's constant, TTT the temperature, and RRR the shunt or load resistance, becoming prominent in high-impedance circuits.[61] Additionally, 1/f (flicker) noise, which follows a 1/fα1/f^\alpha1/fα power spectral density (α≈1\alpha \approx 1α≈1) at low frequencies, stems from surface traps and material defects in semiconductors like InGaAs, significantly impacting baseband signals below 1 kHz.[62]

A key figure of merit for evaluating photodiode performance under noise constraints is the noise-equivalent bandwidth, which quantifies the effective frequency range where the integrated noise equals the shot noise in a 1 Hz band, guiding trade-offs such as increasing reverse bias to enhance speed (by reducing capacitance) at the cost of higher dark current and thus amplified shot noise.[25] For instance, higher bias voltages can achieve bandwidths exceeding 10 GHz but elevate thermal and shot noise densities, necessitating careful circuit design to maintain SNR above 20 dB in communication systems.[63] Mitigation strategies include reducing the active area to minimize junction capacitance (lowering RC limits and thermal noise) and employing transimpedance amplifiers (TIAs), which convert photocurrent to voltage with low-noise op-amps, achieving noise floors as low as 10 fA/√Hz while preserving bandwidth up to several GHz.[64][65]

Desired Photodiode Effects

In photodiodes, photogeneration occurs when incident photons with energy greater than the semiconductor bandgap are absorbed, primarily in the depletion region, creating electron-hole pairs or excitons that contribute to the photocurrent. This process is most efficient in the intrinsic or lightly doped regions where the built-in electric field separates the carriers before recombination, enabling high quantum efficiency in reverse-biased operation.[67]

Carrier collection in photodiodes relies on both drift and diffusion mechanisms, where the strong electric field in the depletion region accelerates minority carriers toward the contacts via drift, while diffusion aids in collecting carriers generated near the edges of the region. This field-assisted transport minimizes recombination losses, achieving quantum yields approaching unity and supporting fast response times in well-designed structures.[67]

Wavelength selectivity in photodiodes arises from the material's bandgap, which determines the cutoff wavelength beyond which absorption is negligible, allowing tailored sensitivity to specific spectral ranges.[68] The relationship is given by the equation

where EgE_gEg​ is the bandgap energy, hhh is Planck's constant, ccc is the speed of light, and λcutoff\lambda_{\text{cutoff}}λcutoff​ is the cutoff wavelength; for example, silicon photodiodes exhibit a sharp response drop around 1100 nm corresponding to Eg≈1.12E_g \approx 1.12Eg​≈1.12 eV.[68] This intrinsic filtering enhances applications requiring discrimination against longer wavelengths, such as in visible-light sensing.

Non-avalanche photoconductive gain in certain photodiodes, particularly those with trap states, results from carrier trapping that prolongs the lifetime of one carrier type, allowing the other to recirculate and amplify the photocurrent beyond unity gain. In wide-bandgap semiconductors like β-Ga₂O₃ used in Schottky photodiodes, this trapping by self-trapped holes enables gains exceeding 50 while maintaining reasonable linearity, beneficial for low-light detection without high-voltage operation.[69]

Thermal effects can positively influence photodiode stability through temperature-compensated bias designs, where controlled heating or circuit adjustments counteract variations in dark current and responsivity, ensuring consistent performance across operating ranges.

Unwanted Photodiode Effects

One significant unwanted effect in photodiodes is dark current, which refers to the small electric current that flows through the device in the complete absence of incident light. This current arises primarily from thermal generation of electron-hole pairs within the semiconductor material, particularly in the depletion region and adjacent quasi-neutral regions. The magnitude of the dark current can be approximated by the formula Idark=qni2A(Lpτp+Lnτn)/NdI_{\text{dark}} = q n_i^2 A \left( \frac{L_p}{\tau_p} + \frac{L_n}{\tau_n} \right) / N_dIdark​=qni2​A(τp​Lp​​+τn​Ln​​)/Nd​, where qqq is the elementary charge, nin_ini​ is the intrinsic carrier concentration, AAA is the junction area, LpL_pLp​ and LnL_nLn​ are the diffusion lengths for holes and electrons, τp\tau_pτp​ and τn\tau_nτn​ are the corresponding minority carrier lifetimes, and NdN_dNd​ is the donor concentration on the n-side (assuming an asymmetric junction).[70] Dark current exhibits strong temperature dependence, typically doubling for every 10°C increase in temperature due to the exponential rise in intrinsic carrier concentration with thermal energy.[23]

At high light intensities, photodiodes can experience saturation, where the output photocurrent no longer increases linearly with input optical power. This occurs due to gain compression mechanisms, such as space charge effects that slow carrier transport, and increased recombination losses in the active region, which reduce the collection efficiency of photogenerated carriers. In photodiode arrays, crosstalk represents another adverse effect, manifesting as unintended signal interference between adjacent elements. Optical crosstalk arises from light scattering or diffraction into neighboring pixels, while electrical crosstalk stems from capacitive coupling or substrate currents that propagate signals laterally through the shared semiconductor structure.[71]

Photodiodes are also susceptible to aging and long-term degradation, which can compromise performance over time. Exposure to radiation, such as in space environments, induces defects in the crystal lattice that increase dark current and reduce responsivity; for instance, silicon-based devices show significant degradation after proton irradiation doses equivalent to years in low-Earth orbit.[72] Humidity accelerates degradation by promoting moisture ingress, leading to corrosion at contacts or passivation layers and elevated leakage currents. These factors contribute to a reduced mean time between failures (MTBF), often estimated using accelerated life testing models that account for environmental stressors to predict operational reliability.

To mitigate these unwanted effects, several strategies are employed. Cooling the photodiode, such as through thermoelectric coolers, suppresses thermal generation and thus lowers dark current by factors of 10 or more per 30-50°C reduction.[73] Guard rings—doped regions surrounding the active junction—help alleviate edge effects by distributing the electric field more uniformly, preventing premature breakdown and reducing surface-related leakage currents.[74] For space applications, where radiation hardness is critical, gallium nitride (GaN)-based photodiodes offer superior tolerance; GaN's wide bandgap and defect-resistant structure provide better performance compared to silicon counterparts.[75]

Optical Sensing and Communication

Photodiodes play a pivotal role in optical sensing and communication systems, where they convert incident light into electrical signals for high-fidelity data transmission and environmental monitoring. In fiber optic receivers, indium gallium arsenide (InGaAs) PIN and avalanche photodiodes (APDs) are widely employed due to their sensitivity in the near-infrared range (900–1700 nm), enabling detection of modulated optical signals in telecommunications infrastructure. These devices support data rates from 10 Gbps to 400 Gbps in Ethernet applications, such as data center interconnects and long-haul networks, by providing low-noise amplification and high bandwidth exceeding 30 GHz. For instance, InGaAs APDs achieve a gain-bandwidth product of over 400 GHz, facilitating error-free operation in dense wavelength-division multiplexing (DWDM) systems. Bit error rates (BER) below 10^{-12} are routinely maintained through forward error correction (FEC) techniques, which correct pre-FEC BERs up to 10^{-3} by adding parity bits, ensuring reliable performance over distances exceeding 100 km without regeneration.

In visible-range applications like barcode scanners and light detection and ranging (LIDAR) systems, silicon photodiodes dominate owing to their cost-effectiveness, fast response times (rise times under 1 ns), and broad sensitivity from 400 nm to 1100 nm. Barcode scanners utilize silicon PIN photodiodes to detect reflected laser pulses from bar patterns, converting intensity variations into electrical signals for decoding at speeds up to thousands of scans per second. In LIDAR, silicon APDs or PIN photodiodes enable pulse detection by measuring the time-of-flight of short laser pulses (typically 10–100 ns duration), achieving sub-millimeter resolution in automotive and surveying applications through precise timing of photon arrival. These photodiodes offer quantum efficiencies above 80% in the visible spectrum, minimizing crosstalk and enabling compact, low-power designs.

Environmental sensing leverages specialized photodiodes for non-contact monitoring of atmospheric conditions. Gallium nitride (GaN) photodiodes excel in ultraviolet (UV) monitors, exhibiting solar-blind operation (cutoff below 365 nm) and high responsivity (up to 0.2 A/W at 254 nm), making them ideal for detecting ozone levels or solar UV index without interference from visible light. These devices operate in photovoltaic mode for low-power, real-time measurement in outdoor stations, with detectivities exceeding 10^{12} Jones for trace UV flux monitoring. For gas detection, photodiodes integrated into absorption spectroscopy systems quantify species like methane or CO2 by measuring light attenuation at specific wavelengths; silicon or InGaAs photodiodes paired with tunable diode lasers achieve parts-per-billion sensitivity through differential detection, reducing noise from ambient light.

Integration of photodiodes with vertical-cavity surface-emitting lasers (VCSELs) in optical transceivers enhances efficiency for short-reach links in 5G and emerging 6G networks. InGaAs or silicon photodiodes are co-packaged with 850 nm or 1310 nm VCSELs in pluggable modules (e.g., QSFP-DD), supporting bidirectional data rates up to 400 Gbps over multimode fiber with power consumption below 10 W per channel. In 5G fronthaul, these transceivers transport digitized radio signals between baseband units and remote radio heads, achieving latencies under 100 μs via time-division duplexing. For 6G prototypes, hybrid integration on silicon photonics platforms incorporates PIN photodiodes for analog fronthaul, enabling terabit-per-second aggregate capacity while meeting power-over-fiber requirements for remote units.

A notable case study is the deployment of photodiodes in submarine fiber optic cables, which underpin over 99% of global internet traffic. At cable landing stations, high-sensitivity InGaAs APDs serve as receivers in terminal equipment, demodulating wavelength-multiplexed signals at rates up to 400 Gbps per channel across transoceanic spans exceeding 10,000 km. While modern repeaters primarily use erbium-doped fiber amplifiers (EDFAs) for all-optical regeneration every 50–80 km, legacy systems and monitoring nodes incorporate PIN/APD pairs for fault detection and performance verification, ensuring BER below 10^{-12} with FEC to support seamless intercontinental connectivity. This infrastructure, exemplified by systems like the MAREA cable (Microsoft/Facebook, 2018), relies on photodiode arrays for multiplexing up to 256 channels, facilitating petabit-scale throughput for cloud services and real-time applications.

Imaging and Scientific Instrumentation

In imaging applications, complementary metal-oxide-semiconductor (CMOS) image sensors employ active pixel sensors (APS) that integrate photodiodes as the primary light-detecting elements, enabling compact and cost-effective designs for consumer cameras such as those in smartphones and digital cameras. Each pixel in a CMOS APS consists of a photodiode paired with one or more transistors for amplification and readout, allowing for on-chip signal processing that reduces the need for external circuitry and improves power efficiency compared to traditional charge-coupled devices (CCDs).[76][77] These sensors dominate consumer markets due to their ability to capture high-resolution images at video rates, with fill factors often exceeding 60% through microlens arrays that direct light onto the photodiode active area.[78]

In scientific instrumentation, photodiode arrays play a crucial role in spectroscopy, particularly in monochromators where silicon (Si) and germanium (Ge) arrays enable precise material analysis across visible and infrared spectra. Si photodiode arrays detect wavelengths from ultraviolet to near-infrared (approximately 200–1100 nm), while Ge variants extend sensitivity into the mid-infrared (up to 1800 nm), facilitating applications like Raman spectroscopy and environmental monitoring by dispersing light through a monochromator and measuring intensity at multiple channels simultaneously.[79][80] These arrays provide high spatial resolution and low crosstalk, essential for resolving spectral lines in chemical composition studies, with quantum efficiencies often reaching 80–90% in optimized Si configurations.[81]

Medical applications leverage photodiodes for non-invasive diagnostics, such as in pulse oximeters operating in transmission mode, where red and near-infrared light passes through tissue like a fingertip, and the transmitted intensity is detected by a photodiode to calculate blood oxygen saturation via the ratio of absorbed light at dual wavelengths (typically 660 nm and 940 nm).[82][83] In fluorescence microscopy, photodiodes, including avalanche variants, detect emitted light from fluorophores excited by lasers, enabling high-sensitivity imaging of cellular structures with sub-micron resolution; for instance, single-photon avalanche diodes (SPADs) integrated into arrays achieve picosecond timing for fluorescence lifetime measurements.[84][85]

In particle physics, silicon photomultipliers (SiPMs)—compact arrays of avalanche photodiodes—serve as key detectors in experiments at CERN, such as the CMS upgrade at the Large Hadron Collider, where they couple to scintillating fibers or crystals to timestamp particle interactions with sub-nanosecond precision and detect low-light yields from high-energy collisions.[86][87] SiPMs offer photon detection efficiencies up to 50%, magnetic field insensitivity, and robustness in radiation environments, outperforming traditional photomultiplier tubes in calorimetry and tracking systems.[88][89]

Photodiode Arrays

Photodiode arrays integrate multiple photodiodes into structured arrangements to enable parallel light detection across spatial dimensions, facilitating applications such as scanning and imaging. Linear one-dimensional (1D) arrays typically consist of 512 to 1024 silicon photodiodes, each approximately 25 μm wide and 2 mm high, arranged in a row for tasks like spectrophotometry where spectral dispersion is captured along a single axis.[94][95] Two-dimensional (2D) arrays expand this to grids with M rows and N columns, often using charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) architectures to form imaging planes in cameras, where horizontally oriented CCD lines or pixel matrices collect light for full-frame capture.[96][97]

Readout mechanisms in photodiode arrays vary by architecture to transfer accumulated charge efficiently. In CCD-based arrays, charge is serially shifted through the array via coupled gates, enabling low-noise transfer but requiring global clocking that can limit speed in large formats.[98] In contrast, CMOS active pixel sensor (APS) arrays incorporate amplifiers and addressing circuitry at each pixel, allowing random access readout for higher speed and integration of on-chip processing, though with potentially higher noise from transistor variability.[77][99] Passive-pixel sensors (PPS), an earlier CMOS variant, rely on shared column gates for readout, offering simpler fabrication but suffering from higher fixed-pattern noise compared to APS due to the lack of per-pixel amplification.[100]

Scalability of photodiode arrays has advanced to support resolutions exceeding 100 megapixels in CMOS implementations, driven by shrinking pixel pitches down to 2-4 μm while maintaining performance.[101] Fill factors, the ratio of active light-sensitive area to total pixel area, routinely surpass 90% through the integration of microlens arrays that focus incident light onto the photodiode, minimizing losses from surrounding circuitry.[102] This enhancement is particularly critical in APS designs where transistor overhead reduces intrinsic photosensitive area.

Key challenges in photodiode arrays include achieving uniformity across elements and preventing blooming. Pixel-to-pixel gain variations are typically held below 10% in CCD arrays through precise fabrication, but CMOS APS can exhibit higher fixed-pattern noise from threshold mismatches, necessitating calibration techniques like correlated double sampling.[103] Blooming, the overflow of excess charge into adjacent pixels during high illumination, is mitigated by anti-blooming gates or vertical overflow drains in CCD structures, limiting spillover to neighboring diodes and preserving image fidelity.[104][105]

Recent advances through 2025 have focused on stacked architectures and back-side illumination (BSI) to overcome traditional limitations. Three-wafer stacked CMOS sensors with BSI photodiodes enable global shutter operation by separating charge collection from readout circuitry, achieving shutter efficiencies over 99.9% without rolling artifacts in high-speed imaging.[106] BSI configurations illuminate the photodiode from the backside, bypassing front-side wiring shadows to boost quantum efficiency (QE) to above 90% in the visible spectrum, particularly for small pixels in megapixel arrays.[107] These innovations, including 2.2 μm pixel stacked BSI global shutter sensors, support compact, high-dynamic-range modules for demanding environments like machine vision.[108]

Pinned Photodiode

The pinned photodiode (PPD) is a buried photodiode structure essential to modern charge-coupled device (CCD) and complementary metal-oxide-semiconductor (CMOS) image sensors, featuring a p⁺/n/p configuration that pins the surface potential to minimize noise and defects.[109] The n-type region serves as the charge collection volume within a p-type epitaxial layer or substrate, while the shallow p⁺ pinning layer at the surface accumulates holes to prevent depletion and isolate the active area from interface traps at the silicon-oxide boundary.[110] This design enables efficient photoelectron accumulation and transfer, distinguishing it from simpler p-n junction photodiodes.[111]

Developed in the early 1980s by Nobukazu Teranishi and colleagues at NEC Corporation, the PPD was initially introduced for interline-transfer CCDs to suppress vertical smear and reduce dark current generation, marking a pivotal advancement over surface-channel photodiodes.[18] Its adaptation to CMOS active pixel sensors in the mid-1990s, led by Eric Fossum at NASA's Jet Propulsion Laboratory, revolutionized consumer and scientific imaging by integrating low-noise detection with on-chip amplification.[109] The structure's physics relies on the pinned potential maintaining a stable electric field, which confines photogenerated electrons in the n-well during exposure while allowing complete depletion for charge transfer.[112]

In typical operation within a four-transistor (4T) CMOS pixel, the PPD integrates photocharge under reverse bias from the substrate contact, with electrons collected in the n-region potential pocket as light absorption creates electron-hole pairs—holes drifting to the p-substrate and electrons to the n-buried layer.[111] Readout involves pulsing the n⁺-implanted transfer gate to form a channel under the gate oxide, rapidly transferring the charge packet to an adjacent floating diffusion node without residual lag, followed by source-follower buffering and correlated double sampling to subtract reset noise.[110] The pinning layer's hole accumulation suppresses trap-assisted generation, yielding dark current densities as low as 0.1–1 e⁻/pixel/s at room temperature in optimized processes.[112]

Key advantages include near-complete charge transfer efficiency (>99.99%), eliminating image lag seen in partially depleted diodes, and inherently low read noise (down to sub-electron levels with CDS), which supports high dynamic range exceeding 70 dB in back-illuminated implementations.[109] The shallow p⁺ layer enhances blue-light sensitivity by reducing absorption losses near the surface, while compatibility with standard CMOS fabrication enables cost-effective scaling to megapixel arrays.[110] Limitations involve potential pinning voltage variability affecting full well capacity (typically 10,000–50,000 electrons), addressed in advanced variants through epitaxial thickness optimization.[111]

Recent innovations extend PPDs to non-traditional substrates, such as thin-film polycrystalline silicon on glass, where a stacked p⁺/i/n/p structure preserves pinning benefits for flexible sensors, achieving conversion gains of 50–100 μV/e⁻ and dark currents below 10 fA/cm² while enabling full charge transfer for noise suppression.[113] Fully depleted PPDs, biased via substrate reverse voltage in high-resistivity silicon, further improve near-infrared quantum efficiency (>80%) and crosstalk reduction in time-of-flight applications.[114] These configurations underscore the PPD's enduring impact, powering over 90% of commercial image sensors as of 2020.[109]