What APD arrays are and why they matter in 2026

Avalanche photodiode (APD) arrays are high-sensitivity photodetector systems that amplify weak light signals through impact ionization, enabling single-photon detection, rapid time-of-flight ranging, and ultra-low-noise imaging across spectral bands from ultraviolet to long-wave infrared. The technology is at an inflection point in 2026, driven by the convergence of LiDAR demand in autonomous systems, next-generation space astronomy requirements, and CMOS integration advances enabling megapixel-scale focal plane arrays.

APD arrays operate by reverse-biasing a semiconductor p-n or heterostructure junction beyond its avalanche threshold, producing internal gain. Within the technology landscape, four primary operating modes are distinguished: linear-mode APD (LmAPD), where gain is proportional to bias; Geiger-mode APD (GAPD), which fires digitally upon single-photon absorption; single-photon avalanche diode (SPAD) arrays, which are CMOS-integrated Geiger-mode devices; and vertical avalanche photodiode (VAPD) configurations optimized for per-pixel time-of-flight circuits.

Material systems represented in the landscape span silicon (Si), indium gallium arsenide on indium phosphide (InGaAs/InP), mercury cadmium telluride (HgCdTe), indium aluminum arsenide (InAlAs), germanium (Ge) and germanium-tin (GeSn), gallium nitride/aluminum nitride (GaN/AlN), and silicon carbide (SiC). Each addresses a distinct spectral window and operating environment, as documented by sources including WIPO patent filings spanning five decades across DE, EP, US, and IL jurisdictions.

Four device architectures driving the APD array field

The four dominant APD array architectures — III-V compound heterostructures, CMOS-integrated SPAD arrays, HgCdTe linear-mode arrays, and wide-bandgap/novel-material devices — each address a distinct performance regime, and understanding their trade-offs is essential for technology and IP strategy.

III-V Compound Heterostructure APDs (InP/InGaAs, InAlAs)

The dominant architecture in active-status patents involves III-V heterostructure designs that spatially separate absorption (InGaAs, InGaAsP, or InAlAs absorber) from multiplication (InP or InAlAs avalanche layer). NTT’s mesa-based architecture — described in multiple active EP patents — uses stacked p-type and low-concentration absorbing layers, a bandgap-gradient layer, field control layers, and an electron transit layer, with a dual-mesa geometry to confine dark current to interior surfaces. Alcatel Lucent’s single-carrier APD (EP, 2019) adds an n+ delta-doped built-in-field layer to accelerate electron injection and reduce device capacitance, targeting high-speed telecommunications.

Source Photonics (Taiwan) demonstrated InAlAs dual multiplication-layer APDs achieving 16–25 GHz bandwidth with 2.5 A/W responsivity. National Central University (Taiwan) showed that triple InAlAs multiplication layers further increase maximum gain to 230 and responsivity to 19.6 A/W compared with dual-layer references.

CMOS-Integrated SPAD and Geiger-Mode APD Arrays

CMOS-integrated SPAD arrays underpin solid-state LiDAR, time-of-flight ranging, fluorescence lifetime imaging (FLIM), and quantum sensing. MIT Lincoln Laboratory’s 20-year program on silicon Geiger-mode APD arrays integrated to all-digital CMOS circuits (reviewed 2016) establishes the lineage. Panasonic Corporation developed a 688×384-pixel VAPD CMOS image sensor achieving 250-meter direct time-of-flight ranging at 10 cm lateral resolution. Jiangnan University fabricated a CMOS APD in 45-nm TSMC technology achieving 8.4 GHz bandwidth at 850 nm, while A*STAR Singapore demonstrated a gain-bandwidth product of 234 GHz at 685 nm with open eye diagrams at 56 Gbps — competitive with infrared counterparts.

HgCdTe Linear-Mode APD Arrays for Low-Background Imaging

Mercury cadmium telluride (HgCdTe) linear-mode APDs exploit the unique band structure of this II-VI alloy to achieve near-noiseless multiplication — excess noise factor approaching unity — in the short-wave infrared (SWIR). The SAPHIRA 320×256 device from Leonardo UK Ltd. was the first dedicated LmAPD product. The dataset documents extension to 512×512 and a 2048×2048 prototype. The University of Hawai’i reported a dark current of approximately 3 electrons per pixel per kilosecond at 50 K in the megapixel prototype (2022). First Light Imaging SAS is co-developing sub-electron noise IR cameras using Leonardo’s 2048×2048 SWIR LmAPD array.

“The University of Hawai’i’s 1-megapixel near-infrared LmAPD prototype achieved a dark current of approximately 3 electrons per pixel per kilosecond at 50 K — a performance level that opens the door to exoplanet spectroscopy and ultra-faint galaxy surveys from ground-based observatories.”

Wide-Bandgap and Novel-Material APDs for Harsh Environments

A fourth cluster covers APDs from wide-bandgap or unconventional materials targeting radiation-hard, high-temperature, UV-sensitive, or structurally novel operation. General Electric Company’s series of IL-jurisdiction patents (2003–2011) covers SiC and GaN APDs for gamma-ray detection in oil well drilling environments requiring 150°C operation and 250-G shock tolerance. Tsinghua University demonstrated a GaN/AlN periodically stacked-structure APD achieving a gain of 10⁴ and an ionization coefficient ratio of 0.05, approaching photomultiplier tube (PMT)-level performance. Samsung Electronics’ 2024 EP patent introduces a graphene tunnel barrier layer to suppress dark current and extend sensitivity across visible and infrared bands simultaneously.

Application domains: from autonomous LiDAR to space astronomy

APD array technology serves six distinct application domains in 2026, each with different performance requirements that map to specific device architectures and material systems.

Autonomous Vehicles and Solid-State LiDAR

SPAD arrays and VAPD CMOS image sensors are the primary technology vectors for automotive ranging. Panasonic’s 688×384 VAPD sensor achieves 250-meter range with sub-photon level signal extraction. Samsung’s graphene APD patent explicitly claims LiDAR system inclusion. Chinese LADAR platforms described in the Southwest Institute / UESTC review applied 32×32–64×64 SPAD arrays to driverless vehicles. InGaAs and Ge/GeSn SWIR APDs are preferred for eye-safe 1550 nm LiDAR operation, as reviewed by the Guangdong Greater Bay Area Institute of Integrated Circuit and System (2023).

Space Astronomy and Adaptive Optics

HgCdTe LmAPD arrays are purpose-built for extreme low-background astronomical applications. The University of Hawai’i and Leonardo UK Ltd. development targets dark current below 1 electron per pixel per kilosecond for exoplanet spectroscopy and ultra-faint galaxy surveys. The Robo-AO adaptive optics demonstration (University of Hawai’i, 2015) achieved 0.73 electrons correlated double sampling read noise in the lab, enabling simultaneous tip-tilt wavefront sensing and high-speed imaging at Palomar Observatory. JPL’s delta-doped silicon arrays (2017) address UV/optical/NIR space telescope applications with wafer-scale post-fabrication processing, as reported by NASA‘s Jet Propulsion Laboratory.

Optical Communications and High-Speed Data Links

InP-based and InAlAs APDs targeting 25–100 Gbps data rates appear in active patents from NTT (EP, 2019–2021) and Alcatel Lucent (EP, 2019). NTT’s inverted p-down design (NTT Device Innovation Center, 2021) enables 100-Gbit/s PAM4 signal handling. A*STAR Singapore’s CMOS-integrated APD for visible-wavelength silicon photonics (2021) reports a gain-bandwidth product of 234 GHz at 685 nm, with open eye diagrams at 56 Gbps.

Defense and Missile Guidance

BAE Systems’ EP patent (2020) covers an algorithm for optimizing APD operating voltage across an array in missile guidance systems, including multi-APD arrays sharing a common operating voltage and accounting for solar background suppression. General Electric’s harsh-environment SiC/GaN APD patents (IL jurisdiction, multiple filings 2003–2011) target downhole oil-well gamma-ray detection requiring 250-G shock tolerance and 150°C operation.

Biomedical and Fluorescence Lifetime Imaging

Silicon APDs with photon trapping nanostructures (University of California, Davis, 2021) demonstrated a 30× gain increase at 850 nm and 50% FWHM pulse-response reduction, targeting time-of-flight and FLIM applications. Columbia University’s 72×60 angle-sensitive SPAD (A-SPAD) array in 180-nm CMOS (2016) enables lens-less 3D FLIM down to micrometer-scale resolution. These developments are relevant to the broader single-photon detection research community tracked by Nature Photonics.

Particle Physics and High-Energy Astrophysics

Geiger-mode APD prototype cameras were deployed for very-high-energy gamma-ray astronomy (ETH Zurich, 2009). Large-area APDs (LAAPDs, Advanced Photonix) are characterized for soft X-ray detection down to a 0.5 keV threshold and are being studied as photomultiplier-tube replacements in experiments such as the Cherenkov Telescope Array, a project coordinated under standards frameworks from IEEE.

Patent assignees and geographic concentration in the APD array IP landscape

NTT holds the densest active patent position in the dataset, with four active EP patents (2019–2021) all focused on InP-based high-speed APD structures for telecommunications — making freedom-to-operate analysis against the NTT EP portfolio essential before product launch in European markets.

Sony Semiconductor Solutions Corporation filed one active EP patent in 2024 targeting SPAD-mode sensors with precise impurity concentration engineering, representing Japan’s consumer/imaging-sector APD push. Samsung Electronics filed one active EP patent in 2024 introducing a graphene-barrier APD architecture with LiDAR application claims — representing Korea’s emerging position in the field. BAE Systems Information and Electronic Systems Integration Inc. holds one active EP patent (2020) covering array-level voltage optimization for defense APD systems. Dephan Limited Liability Company holds two active IL patents (2019, 2021) on a substrate-integrated APD architecture based on PCT/RU2016/000708, indicating Russian-origin IP entering commercial jurisdictions.

Geographic concentration in research literature is notably broad. Institution activity spans the US (MIT Lincoln Laboratory, JPL, UC Davis, University of Hawai’i, Columbia University), UK (Leonardo UK Ltd.), France (First Light Imaging SAS, Alcatel Lucent), Singapore (A*STAR), Japan (NTT, Panasonic), Taiwan (Source Photonics, National Central University, Jiangnan University), and China (Southwest Institute of Technical Physics, UESTC, Tsinghua University, Guangdong Greater Bay Area Institute). This distribution suggests the field is not concentrated in a single national ecosystem — a pattern consistent with global semiconductor innovation trends tracked by OECD science and technology indicators.

Five emerging directions reshaping avalanche photodiode array technology

The most recent filings and publications (2021–2024) in the dataset point toward five directions that will define the next generation of APD array products and IP positions.

1. Megapixel LmAPD Formats for Space and Adaptive Optics

The scaling of HgCdTe LmAPD arrays to 1-megapixel (and beyond toward 2048×2048) represents the leading format frontier. The University of Hawai’i’s prototype (2022) targeting fewer than 1 electron per pixel per kilosecond dark current, and Leonardo UK’s 2048×2048 SWIR LmAPD under active development with First Light Imaging (2022), point toward near-term availability of space-qualified megapixel APD arrays for wavefront sensing, exoplanet characterization, and ground-based adaptive optics.

2. Graphene and Novel 2D-Material Tunnel Barriers

Samsung Electronics’ 2024 EP patent explicitly incorporates a graphene layer as a tunnel barrier between collector and emitter layers in an APD pixel designed for both visible and infrared response and LiDAR integration. This is the most structurally novel approach in the dataset and signals that 2D-material integration is moving from research to patent-stage product development. IP strategists should track continuation filings and divisional applications from this family closely.

3. Multi-Multiplication-Layer InAlAs APDs for Coherent Communications

The progression from single to dual to triple InAlAs multiplication layers — demonstrated by Source Photonics (2021) and National Central University (2022) — reduces breakdown voltage, increases gain, and relaxes the responsivity-saturation tradeoff critical for coherent optical receivers. This trajectory suggests commercial InAlAs multi-layer APDs for 400G and 800G coherent links are imminent.

4. Photon-Trapping Nanostructures for Silicon APD Arrays

UC Davis’s 2021 demonstration of silicon APDs with photon-trapping nanostructures achieved greater than 60% absorption efficiency at 850 nm and greater than 20× gain enhancement, indicating that structural photon management is emerging as an alternative to material substitution for extending the sensitivity and speed of CMOS-compatible APD arrays into near-IR bands relevant to biomedical and automotive applications.

5. SWIR APD Extension via Ge/GeSn and InGaAs for Eye-Safe LiDAR

The 2023 review by the Guangdong Greater Bay Area Institute documents the maturation of Ge, GeSn, and InGaAs APDs targeting the 1300–1550 nm SWIR window for LiDAR applications. The emergence of GeSn alloys as a CMOS-compatible SWIR absorber is particularly significant, as it may enable fully monolithic APD array fabrication on silicon substrates — a development with direct implications for cost reduction in high-volume automotive LiDAR, consistent with semiconductor roadmaps published by the Semiconductor Industry Association.

“Samsung’s 2024 graphene-barrier APD patent is an early signal worth monitoring: if 2D-material tunnel barriers prove manufacturable at scale, they could disrupt both the III-V APD and HgCdTe LmAPD supply chains by enabling broader-spectrum, lower-dark-current operation on standard substrates.”

Strategic implications for R&D and IP teams in 2026

Five strategic conclusions follow directly from the patent and literature evidence in this landscape, each with actionable consequences for R&D investment decisions and IP portfolio management.

• CMOS integration is the primary commercialization enabler. The convergence of SPAD/GAPD arrays with standard 45-nm to 180-nm CMOS processes eliminates the need for hybrid flip-chip bonding and reduces cost at scale. R&D teams targeting volume markets (automotive LiDAR, consumer ToF, biomedical FLIM) should prioritize CMOS-compatible device architectures over maximum-performance compound semiconductor alternatives.
• NTT’s EP portfolio is the primary III-V freedom-to-operate risk. With four active EP patents on mesa-based InP dark-current suppression, NTT represents a significant IP obstacle for companies seeking to commercialize InP/InGaAs APDs in European markets. Freedom-to-operate analysis against the NTT EP portfolio is essential before product launch.
• Samsung’s graphene-barrier APD patent (2024) warrants close monitoring. If 2D-material tunnel barriers prove manufacturable at scale, they could disrupt both the III-V APD and HgCdTe LmAPD supply chains. IP strategists should track continuation filings and divisional applications from this family.
• HgCdTe LmAPD arrays represent a near-monopoly opportunity window. With Leonardo UK Ltd. as the only named supplier of dedicated LmAPD products in this dataset and academic groups (University of Hawai’i, First Light Imaging) as the primary integrators, the commercial supply chain for megapixel-format LmAPDs for astronomy and adaptive optics is fragile. New entrants with HgCdTe or alternative low-excess-noise materials could command significant pricing power.
• Wide-bandgap APDs represent a white-space IP opportunity. GE’s lapsed SiC/GaN APD portfolio (IL jurisdiction, last filing 2011) and BAE Systems’ active EP voltage-optimization patent suggest the harsh-environment APD space has seen reduced patent activity since 2015 — potentially indicating a white-space opportunity for new IP claiming SiC/GaN APD array architectures updated for modern CMOS readout integration.