Linear-Mode Gain HgCdTe Avalanche Photodiodes for Weak-Target Spaceborne Photonic System

Abstract

Spectroscopic observations of Earth-like exoplanets and ultra-faint galaxies–top scientific priorities for the coming decades–involve measuring broadband signals at rates of only a few photons per square meter per hour. This imposes exceptional requirements on the detector performance, necessitating dark currents below 1 e⁻/pixel/kilo second, read noise under 1 e⁻/pixel/frame, and the ability to handle large-format arrays–capabilities that are not yet met by most existing infrared detectors. In addition, spaceborne LiDAR systems require photodetectors with exceptional sensitivity, compact size, low power consumption, and multi-channel capability to facilitate long-range range finding, topographic mapping, and active spectroscopy without increasing the instrument burden. MCT Avalanche photodiodes arrays offer high internal gain, pixelation, and photon-counting performance across SW to MW wavelengths needed for multi-beam and multi-wavelength measurements, marking them as a critical enabling technology for next-generation planetary and Earth science LiDAR missions. This work reports the latest progress in developing Hg_1−xCd_xTe linear-mode e-APDs at premier industrial research institutions, including relevant experimental data, simulations and major project planning. Related studies are summarized to demonstrate the practical and iterative approach for device fabrication, which have a transformative impact on the evolution of this discipline.

1. Introduction

Conventional infrared detectors suffer from limitations due to read noise and integration capacitor size, particularly in photon-starved conditions. These constraints reduce detection sensitivity and dynamic range, posing significant challenges for applications such as high-speed wavefront sensing, LiDAR laser-gated imaging and faint galaxy spectroscopy) [1,2,3,4,5,6]. Recent advances in infrared detector technology, especially the development of HgCdTe avalanche photodiodes (APDs) operating in linear mode (LM), offer effective solutions to these issues. Broad spectral sensitivity (0.9–4.3 µm) Hg_1−xCd_xTe LM-APDs are essential because of their ability to enable infrared detection through internal avalanche gain amplification, meeting the mission-critical demands of aerospace, defense, and scientific domains [7,8,9,10,11,12]. Additionally, their high quantum efficiency, minimal dark current, and linear response maintain sufficient sensitivity, enabling continuous, high-speed operation even under low photon flux. For applications requiring time-resolved or continuous measurements, linear-mode detectors can be more suitable because they can handle photon fluxes without the recovery dead-time associated with Geiger-mode detectors. LM HgCdTe infrared detectors offer distinct advantages over Geiger-mode SPADs with the demands of LiDAR systems, including larger dynamic range, proportional output to photon count, and absence of dead time after-pulsing [13,14]. Advanced LiDAR applications including carbon dioxide and methane monitoring, terrain mapping of Mars, the Moon, and asteroids, and water ice detection are key priorities supported by the NASA Earth Science Technology Office (ESTO) and Science Mission Directorate (SMD) [15,16,17,18,19,20,21,22]. The LM HgCdTe detectors developed through the collaboration between the University of Hawai’i and Leonardo (formerly Selex) are specifically optimized for ultra-low background astronomical scenarios, and cosmological observations requiring stringent few-photon sensitivity with long exposures [2,23,24]. The latest megapixel APD results demonstrate excellent performance, with dark current below 1 × 10⁻⁴ electrons/pixel/second and read noise dropping by 30% per volt of bias, averaging down to 1 e⁻/pixel/frames [8]. Within the framework of the HOLDON H2020 European project, CEA-Leti led the development of a customized HgCdTe avalanche photodiode and CMOS readout circuit, working alongside industry partners and end-users to design, test, and prepare the detector for integration into space-based LiDAR systems targeting greenhouse gas detection [25].

MCT avalanche photodiodes operate via an electron-initiated avalanche gain mechanism, which becomes particularly effective at lower cadmium concentrations (x < 0.65), where the ionization coefficient ratio (k ≪ 1) favors electron multiplication [26]. Electrons, with their low effective mass and high mobility, can efficiently gain energy from modest electric fields and undergo impact ionization, whereas heavy holes exhibit low mobility and strong phonon scattering, making them less effective in this process. The absence of low-lying secondary minima in the HgCdTe conduction band further allows electrons to reach high energy levels, increasing the ionization probability [27,28,29,30,31,32]. Several institutions, including the DRS, AIM, and CEA/Leti, have developed various device structures to exploit this regime. These devices feature a p-type active region and a central n-type multiplication layer that, under increasing reverse bias, become fully depleted to enable avalanche multiplication via high-energy electrons. These material and device-level advances have laid the foundation for ongoing efforts to develop low-threshold APDs, which enables more deterministic avalanche processes at lower bias voltages. The pursuit of lower excess noise and improved gain-bandwidth performance has motivated the exploration of advanced materials exhibiting advantageous impact ionization properties [33,34,35,36,37].

In this review, we summarize the current state and future advancements of IR-based HgCdTe APDs. In Section 2, we briefly introduce the principles, representations, and technical strategies for HgCdTe APDs. In Section 3, we organize six main topics to review recent progress in the state-of-art strategies and architectures for high-performance HgCdTe APDs. This paper explores key opportunities and challenges in the field. It aims to provide a comprehensive overview of infrared technology utilizing APD arrays, with the objective of fostering further research in this area.

2. Fundamentals of MCT Avalanche Photodiodes

MCT-APDs are designed to sustain high electric fields using doped intrinsic layers, and photoexcited carriers are driven by bias into multiplication layers, where amplification occurs through charge scattering and impact ionization [38,39]. Proper modelling of the electric field distribution is crucial for optimal performance. Linear-mode detectors operate at a lower gain, which allows the photon-induced current to be proportional to the number of incident photons, rather than just triggering an all-or-nothing avalanche event. The advantage of linear mode is that it provides greater linearity and allows for more detailed measurement of photon intensities, which can be important in applications where photon counting needs to be precise and quantitative [3,40].

2.1. Multiplication Effect and Internal Gain Mechanism

The avalanche process in narrow bandgap semiconductors HgCdTe is complex due to strong alloy scattering, requiring specialized theoretical models [41,42,43,44]. The classical McIntyre model, which defines impact ionization by local electric fields, was refined with a history-dependent avalanche model incorporating dead-space effects [31]. The remarkable silicon-compatible thermal expansion of HgCdTe, and its high dielectric constant, superior electron mobility, and efficient electron-hole separation make it ideal for low-noise APD fabrication and large-scale integration with silicon readout circuits [28,45]. Heavy holes process a large effective mass, resulting in slow migration and significant scattering with optical phonons, which causes most of their acquired energy to dissipate. In contrast, electrons exhibit a remarkably small effective mass, high mobility, and weak interaction with optical phonons, allowing them to gain energy from the electric field efficiently [46,47].

For the planar n⁺-n⁻-p HgCdTe e-APD, the n⁺ layer is formed via boron ion implantation into HgCdTe epitaxial film. The high doping concentration in the n⁺ layer also helps form a sharp junction with the underlying n⁻ layer, contributing to the device’s electric field profile. The n⁻ layer is created through a post-implantation annealing process, which serves as the multiplication layer and plays a critical role in achieving avalanche gain. Under reverse bias, it becomes fully depleted and supports a high electric field necessary to initiate impact ionization events, as shown in Figure 1. The doping profile of this region is crucial for controlling the gain, threshold voltage, and excess noise factor of the APDs.

2.2. Representation and Performance Metrics

A set of key performance metrics was established to effectively evaluate and compare the performance of fabricated APD photodetectors across various architectures and operating conditions. To optimize detection sensitivity, designs that focus on carrier transport mechanisms with minimal energy dissipation and scattering can be implemented, facilitating almost instantaneous carrier movement. Numerical model for MWIR HgCdTe APDs was developed generally using SILVACO TCAD software, which utilizes the Okuto–Crowell model [48]. Various recombination and generation mechanisms are considered in the transport model, including SRH recombination [49], trap-assisted tunneling (TAT) [50], band-to-band tunneling (BBT) [51], avalanche generation [52], Auger recombination [53], and radiative recombination [54], which are summarized in

Table 1. Comprehensive analysis of dark current components [55].

Transport Models	Equation
Drift–diffusion model	$\nabla ·\overrightarrow{J_n}=q{R}_{net,n}+q{\displaystyle \frac{\partial n}{\partial t}},-\nabla ·\overrightarrow{J_p}=q{R}_{net,p}+q{\displaystyle \frac{\partial p}{\partial t}}$
SRH recombination and TAT	$R_{net}^{SRH\&TAT}={\displaystyle \frac{pn-n_{i,eff}^2}{{\tau }_p\left(n+n_1\right)+{\tau }_n\left(p+p_1\right)}}$$,n_1=n_{i,eff}\exp \left({\displaystyle \frac{E_{trap}}{KT}}\right),p_1=n_{i,eff}\exp \left({\displaystyle \frac{{-E}_{trap}}{KT}}\right)$
BBT generation	$G^{BBT}=A_{BBT}{E}^P\exp \left({\displaystyle \frac{B_{BBT}}{E}}\right),A_{BBT}=-{\displaystyle \frac{q^2\sqrt{2m_e^{*}}}{\pi h^2\sqrt{E_g}}},B_{BBT}={\displaystyle \frac{{\pi }^2\sqrt{\frac{m_e^{*}}{2}}{E}_g^{\frac{3}{2}}}{qh}}$
Avalanche generation	$G^{avalanche}=a_{n}n{v}_n+a_{p}p{v}_p,\alpha \left(E\right)=a{E}^c\mathrm{e}\mathrm{x}\mathrm{p}({(-{\displaystyle \frac{b}{E}})}^2)$
Auger recombination	$R_{net}^{auger}=A_{1}n\left(pn-n_{i,eff}^2\right)+A_{7}p\left(pn-n_{i,eff}^2\right)$
Radiative recombination	$R_{net}^{radiative}=C\left(pn-n_{i,eff}^2\right)$

. The dark current transport is mainly influenced by the concentration of doped carriers. At low bias, SRH and TAT mechanisms dominate the dark current, whereas at high bias, the BBT and avalanche processes become the primary contributors.

The doping profiles of the junctions were estimated through capacitance–voltage (C-V) measurements. The junction width (W) was calculated using Formula (1), which involves the junction area (A), capacitance (C), vacuum dielectric constant (ε₀), and dielectric constant of HgCdTe (ε_s). The effective doping concentration was determined by the separate Equation (2).

$$W={\displaystyle \frac{A{\mathsf{\epsilon }}_0{\mathsf{\epsilon }}_s}{C}}$$

(1)

$$N^{*}\left(W\right)={\displaystyle \frac{C^3}{q{A}^2{\mathsf{\epsilon }}_0{\mathsf{\epsilon }}_s}}{\left[{\displaystyle \frac{dC}{dV}}\right]}^{-1}=\left[{\displaystyle \frac{1}{P\left(W\right)}}+{\displaystyle \frac{1}{N\left(W\right)}}\right]$$

(2)

The multiplication gain (M), and gain-normalized dark current density (GNDCD) were derived by calculating the ratio of the high-bias current to the zero-bias current, considering both photon $I_{\mathit{photon}}$ and $I_{\mathit{dark}}$ dark currents.

$$M(V)={\displaystyle \frac{I_{\mathit{photon}}\left(V\right)-I_{\mathit{dark}}\left(V\right)}{I_{\mathit{photon}}(V=0)-I_{\mathit{dark}}(V=0)}}$$

(3)

$$\mathit{GNDCD}={\displaystyle \frac{I_{\mathit{dark}}\left(V\right)}{M·A}}$$

(4)

Evaluating the excess noise factor (F(M)) in an APD is essential for determining the ultimate performance of photoelectric systems, as it helps assess the signal-to-noise ratio degradation during amplification. Multiplication noise was measured as a function of reverse bias, and the F(M) was estimated using a specific formula [56]. From the signal-to-noise ratio perspective, the F(M) is the ratio of signal-to-noise ratio of the device input to its output signal-to-noise ratio. The APD amplifies all input signals through avalanche multiplication, but this process introduces excess noise owing to the randomness of impact ionization.

$$F={\displaystyle \frac{{\mathit{SNR}}_{\mathit{in}}}{{\mathit{SNR}}_{\mathit{out}}}}={\displaystyle \frac{{\mathit{SNR}}_{M=1}}{{\mathit{SNR}}_M}}=({\displaystyle \frac{I_0/\sqrt{2q{I}_0}}{I_{0}M/\sqrt{2q{I}_0{M}^{2}F}}}{)}^2=({\displaystyle \frac{V_{\mathit{noise}\left(M\right)}{C}_M{T}_{\mathit{int}(M=1)}}{{MV}_{\mathit{noise}(M=1)}{C}_{M=1}{T}_{\mathit{int}\left(M\right)}}}{)}^2$$

(5)

Gain uniformity at a high reverse bias is crucial for linear mode APD operation with low excess noise and dark current at higher gains, and improving the gain uniformity of HgCdTe APDs is essential for low excess noise applications. The excess noise factor (F(M)), determined by the randomness of the multiplication process with the standard deviation ${\mathsf{\sigma }}_M$, and for each pixel in an array, was measured through gain fluctuations across multiple frames, typically less than 1.5.

$$F=1+{\displaystyle \frac{{\mathsf{\sigma }}_M^2}{M^2}}$$

(6)

The NEPh performance is determined by the combined performance of the APD detector array and CMOS ROIC, which is defined as the number of integrated photons that produce a signal equal to the quiescent noise. The NEPh in the APD FPA is calculated using factors such as the noise equivalent power (NEP), integration time (T_int), pixel area (A_d), normalized detectivity (D*), background flux generated current density ($J_{\mathit{\phi }}$), and noise equivalent input of the readout expressed in terms of RMS electrons (N_ee) [57].

$$\mathit{NEPh}={\displaystyle \frac{\mathit{NEP}·T_{\mathit{int}}}{h·\frac{c}{\mathsf{\lambda }}}}={\displaystyle \frac{\mathsf{\lambda }·\sqrt{A_d·T_{\mathit{int}}}}{h·c·D^{*}}}={\displaystyle \frac{\sqrt{{[M}^2·F(M)·(\mathit{GNDCD}+J_{\mathit{\phi }})·A_d·\frac{{\mathsf{\tau }}_{\mathit{gate}}}{q}]+N_{ee}^2}}{M·\mathsf{\eta }·{\mathit{FF}}_{ug}}}$$

(7)

The external quantum efficiency (QE)-$\mathsf{\eta }\left(\mathsf{\lambda }\right)$, or incident photon-to-current efficiency, is the ratio of extracted electrons to incident photons. It is often expressed in terms of responsivity $R\left(\mathsf{\lambda }\right)$, which relates the output current or voltage to the input photon power, depending on the measurement method.

$$\mathsf{\eta }(\mathsf{\lambda })=R(\mathsf{\lambda }){\displaystyle \frac{hc}{\mathsf{\lambda }q}}$$

(8)

2.3. Primary Approaches and Technical Strategies

Three main types of e-APD technologies are currently reported in the literature: planar diffused and via-hole processes, which are homojunctions with constant bandgap, and mesa heterojunctions, where doping and bandgap can vary throughout the device structure. Leonardo in Southampton has developed metal–organic vapor phase epitaxy (MOVPE) for HgCdTe growth on GaAs substrates for mesa heterojunctions. MOVPE offers full control over the bandgap and doping profiles, allowing optimization of the absorber, p-n junction, and multiplication regions independently (Figure 2a). Each pixel was electrically separated by a mesa slot that extended through the absorber, preventing lateral charge collection and blooming. Bandgap widening around the edges of absorber helps isolate carriers from surface states and reduces junction currents where the junction meets the sidewall [58]. MOVPE was initially thought to be unsuitable for e-APDs because of potential misfit dislocations; however, these issues can be mitigated through bandgap engineering and optimized device design. The technology also benefits from reduced optical crosstalk and inter-capacitance, making it ideal for astronomy applications. The CdTe layer coats the sidewalls of the mesa to reduce junction currents, enhance stability, and improve the reverse-bias breakdown performance. This hetero-passivation in HgCdTe significantly enhances the performance and longevity of e-APDs in high-temperature environments. By combining heavy doping buffer, proper metal selection, and possible alloying or interfacial engineering, the low-resistance ohmic contacts can be effectively achieved to promote adhesion and reduce interface resistance.

The HgCdTe High-Density Vertically Integrated Photodiode (HDVIP) structure, designed for front-side illumination, features a p-around-n junction. Fabrication began with a 6–7 µm thick p-type, Cu-doped thin film grown via liquid phase epitaxy (LPE) or molecular beam epitaxy (MBE), with the Hg fraction controlling the bandgap and cutoff wavelength (Figure 2b). Both sides of the material were passivated with inter-diffused CdTe to reduce surface recombination. The double-sided inter-diffused p-type structure was then attached to ROIC though epoxy, where vias were etched to form n-type cylindrical regions around them, followed by metallization for electrical connections [59]. During the via etching process, mercury interstitials are introduced, filling vacancies within the p-type HgCdTe material. This redistribution of vacancies causes the Cu acceptor dopants to migrate and accumulate in regions of higher vacancy concentration, resulting in the formation of a localized n-type cylindrical region around each via.

The planar n⁺/n⁻/p HgCdTe APDs produced by Leti and Sofradir were fabricated using LPE-grown on CdZnTe substrates [60]. These devices feature a homo-junction with the n⁻ region created by converting a localized surface area to n-type, as shown in Figure 2c. The p-type absorber layer was formed with vacancy doping, and the shallow n⁺ region was created by boron ion implantation. The depleted n⁻ multiplication region was formed through annealing to reduce Hg vacancies, achieving avalanche gain multiplication. The thickness of the layers and pixel pitch influence the diffusion volume, which affects dark current at high temperatures. At low temperatures, dark current arises mainly in the multiplication region through trap and tunneling mechanisms, which are enhanced by the electric field.

3. Progress in LM-APD Technology for HgCdTe Infrared Detectors

3.1. Leonardo UK Infrared Sensors Advancing Precision in Scientific Research

Leonardo UK infrared sensors were designed to address the demanding requirements of scientific applications. These sensors are particularly suited to astronomy and Earth observation, where extreme operating conditions and low photon arrival rates are common challenges. By employing advanced technologies, such as bandgap-engineered HgCdTe, produced via MOVPE, these sensors achieve superior performance. Critical innovations include reducing dark current and enhancing pixel separation, leading to top-tier resolution, minimal crosstalk, and low inter-pixel capacitance, which makes Leonardo sensors well-suited for high-precision scientific research [61,62,63].

3.1.1. 320 × 256 e-APD Infrared Saphira Detector

The SAPHIRA detector, originally developed for fringe-tracking at the European Southern Observatory, is now the leading choice for NIR wavefront sensing in adaptive optics. Manufactured by Leonardo (formerly Selex), it has been evaluated for its ability to count NIR photons in collaboration with the University of Hawai’i Institute for Astronomy (UH-IfA) [58]. The C-RED One is an ultra-low noise infrared camera built by First Light Imaging, featuring the Saphira detector designed by Leonardo UK. It is tailored for high-speed infrared applications, offering exceptional photon sensitivity (<1 photon rms) with Fowler sampling and non-destructive readout at speeds over 10 K frames/s. The camera was equipped with an HgCdTe APD array, suitable for 1–2.5 µm wavelengths. C-RED One operates as an autonomous, plug-and-play system with a user-friendly interface that is ideal for extreme and remote locations. It uses a self-managed sealed vacuum and cryogenic cooling via an integrated pulse tube, which requires no human intervention [64]. The QE peaks at nearly 80%, and results show that the RMS noise is 0.60 e⁻ at gain 50 and 1720 fps. The detector has undergone several improvements, including the removal of bias oscillation and damping of detector vibrations to 0.3 µm RMS jitter. The background noise was reduced using ME1001 ROIC, which eliminated the intrinsic glow source. The noise is now 30–40 e⁻/s at gain 1, and the total noise with a cap is approximately 0.7 e⁻ RMS at gain 50. Additionally, the camera now has minimal or no defective pixels, benefitting from the high-quality MCT-MOVPE growth.

Assessments of Leonardo 320 $\times$ 256/24 µm pixel pitch APD arrays revealed that the noise floor is below the signal level, and 2D data captured with a 500 ns gate time can function within maximum bias and gain conditions [65]. The noise increased slightly around 11–12 V, but significantly increased at reverse bias >16 V, likely due to trap-assisted tunnel currents in the APD structure. This led to a higher defect rate, with defective pixels rising from <0.1% at reverse bias <16 V to 4.5% at 17.7 V, corresponding to median APD array gain of M = 466. The signal uniformity was observed to be >90% in the central region of the array, with the central pixels showing better consistency. The perimeter pixels exhibited an increased response, leading to an asymmetric signal histogram. The results highlight that the array exhibits good signal uniformity in the central region, but experiences increased defects and noise at higher reverse bias voltages. Development plans aim to address interface data and configuration constraints to further evaluate the FPA performance under various conditions, including different gains, laser pulse widths and powers, gate times, frame rates, and higher readout speeds in 2D/3D imaging modes.

3.1.2. Initial Characterization of ME1120 ROIC for 512 × 512 Pixel SAPHIRA e-APD Array

The new 512 $\times$ 512/24 µm pixel large SAPHIRA infrared e-APD array, an extension of the 320 $\times$ 256 SAPHIRA device, was developed by Leonardo in collaboration with Max Planck Institute for Extraterrestial Physics (MPE), National Research Council Canada (NRC), and European Southern Observatory (ESO) [66,67]. The ME1120 ROIC, tested with the ESO NGC controller, supports advanced readout modes, such as global or rolling reset and rolling or read-reset-read per row. With a maximum frame rate of 4000 fps, the ROIC can handle rapid imaging without significant performance degradation, exhibiting 1 e⁻ glow per 1000 frames. The device offers multiple windowing options and precise avalanche gain control, with a maximum gain of up to 300, using single negative power supply up to 18 V. The infrared sensor operates efficiently across a broad waveband of 0.8–2.5 µm (partial 3.5 µm) and achieves a typical read noise of 0.5 e⁻ at a gain of 80. Additionally, the device maintains low noise figure of less than 1.2 and operates effectively within a temperature range of 40 to 140 K, making it well-suited for high-performance infrared imaging applications in demanding environments. Process improvements aim to reduce dark current and enhance the APD gain, with a new diode structure featuring a ramped bandgap that is expected to reduce BBT and improve performance. These advancements, including sub-electron readout noise for long integration times, are targeted for use in the GRAVITY+ science spectrometer and other ELT applications, where high-speed, low-noise detectors are critical for photon-starved science cases.

3.1.3. 1 Megapixel NIR APD Array: Ultra-Low Background Space Astronomy Tests

Leonardo Corporation, in collaboration with ESO and the University of Hawai’i, structured the program as three primary development stages in

Table 2. Comparison of characteristics of SAPHIRA detectors versus 1 k × 1 k LM-APDs in both engineering and science-grade versions [68].

Parameter	SAPHIRA	ME1070 (Engineering-Grade, Already Tested)	ME1070 (Science-Grade, This Work)
Pixel size	24 µm	15 µm	15 µm
Format	320 × 256 pixs2	1024 × 1024 pixs2	1024 × 1024 pixs2
Max frame rate	1000 Hz	∼10 Hz	∼10 Hz
Reference pixel	No	Yes	Yes
Number of video outputs	32	16	16
Bandgap structure	Widened	Widened	Widened and graded
Onset of tunneling current	8 volts	8 volts	13 volts (est.)

[2]: Fabricate and validate 1 k × 1 k with SAPHIRA-equivalent bandgaps, smaller pixels, and reference pixels to demonstrate large-format manufacturability (without targeting maximum sensitivity). Produce and evaluate larger 1 k × 1 k arrays featuring a compositionally graded bandgap designed to simultaneously achieve both low dark current and low read noise. Subject the science-grade detectors to space-relevant radiation environments to confirm their robustness for future space missions [68]. Thorne et al. demonstrate the development of scientific-grade, megapixel-format linear-mode APD specifically optimized for low-background shortwave infrared astronomy in the 1–2.4 µm range, and present newly acquired observational data from initial on-sky measurements. The most recent prototypes exhibit the potential for photon-number resolving operation: dark currents below 1 × 10⁻⁴ e⁻/pixel/s, read noise that decreases by 30% with each volt of bias—falling under 1 e⁻/pixel/frame in correlated double sampling, and averaging down to approximately 0.3 e⁻/pixel/frame using multiple non-destructive reads [2,68,69,70].

3.1.4. Sub-Electron Noise Infrared Camera Using Leonardo 2 K × 2 K SWIR LmAPD Array

Recently, larger Lm-APD arrays funded by the ESA have been under development for low-background astronomy and adaptive optics, including arrays with sizes up to (2 K $\times$ 2 K/15 µm), with high frame rates and low noise/flux performance [71]. Leonardo is developing a new ME1130 ROIC to drive future 2 K $\times$ 2 K Lm-APD detectors as shown in Figure 3b. The key features of ME1130 include a 2048 $\times$ 2048 format, 15 µm pixel pitch, low glow design, configurable multiple windows, selectable 16, 8, or 4 analog video outputs, non-destructive readout, and low intrinsic noise.

3.2. Performance Enhancements in e-APDs at Leonardo DRS for Spaceborne LiDAR

Mercury cadmium telluride LM e-APDs overcome longstanding limitations in single-photon detection across the SW to MW infrared spectrum (0.9–4.3 µm). This technology has unlocked new NASA spaceborne LiDAR capabilities, supported by ESTO and SMD. NASA Goddard Space Flight Center has partnered with Leonardo DRS Electro-Optical Infrared Systems to develop HgCdTe APD arrays for spaceborne LiDAR receivers.

3.2.1. HgCdTe APD Arrays for Surface Elevation and Atmospheric LiDAR

Leonardo DRS has developed 4 × 4-pixel HgCdTe APD arrays—each 80 × 80 µm with four parallel APDs per pixel and a surrounding guard band—using cylindrical HDVIP technology. Featuring an 8 MHz TIA bandwidth (20 ns rise/50 ns fall) with CTIA readout capable of femtoampere resolution, these arrays deliver single-photon sensitivity across visible to 4.4 µm wavelengths (QE >90%), a 75% fill factor, no inter-pixel dead space, and a linearly increased dynamic range. The 4 × 4-pixel HgCdTe APD arrays have powered airborne IPDA lidar missions at 1.57 µm (CO₂, 2014–2017) and 1.65 µm (CH₄, 2017), as well as the 2.05 µm CO₂ system. These arrays have also been integrated into prototype planetary lidars—measuring Martian wind and backscatter profiles, lunar water-ice at 3 µm in SpIRRL, and small-body surface composition with a 2 × 8 LMPC IDCA for asteroid and comet missions as summarized in

Table 3. Characteristics of the 4 × 4-pixel and 2 × 8-pixel HgCdTe APD arrays [76].

Parameters	4 × 4-Pixel HgCdTe APD Arrays	2 × 8-Pixel HgCdTe APD Arrays
Quantum efficiency	>90%, 0.9 to 4.4 µm	>90%, 0.9 to 4.4 µm
APD gain	1 to 900, APD bias (0–12 V)	1 to1900, APD bias (0–12 V)
Excess noise factor	1.05	1.15
Dark current	<0.5 pA/pixel	<8 fA (50,000 electrons/s) per pixel
Responsivity	>2 × 109 V/W	1.0 to 1.5 × 109 V/W with amplifiers
Electrical bandwidth	8 MHz	50 MHz
NEP at 1.55 µm	<0.5 fW/Hz1/2/pixel	<0.2 fW/Hz1/2/pixel
Pixel size	80 × 80 µm	64 × 64 µm
Pixel pitch	80 µm	64 µm
Photographs

[22,72,73,74]. Since 2012, the Leonardo DRS has advanced its 2 × 8-pixel LMPC HgCdTe APD array, initially demonstrating a 4.4 µm-cutoff HDVIP device with 64 × 64 µm pixels, 22 µm active regions, avalanche gains of 500 at 12 V bias, 6–9 ns pulse widths, and 60% photon-detection efficiency at a false-event rate of 10⁶ counts/s. Under NASA GSFC ESTO ACT program, dark counts were reduced tenfold with a metal glow-shield and gains rose to 1900, leading to two IDCA modules with improved cold-shielding and low dark noise as shown in Figure 4. These units exhibited consistent PDE/FER at 1.03 µm and 1.55 µm, a usable dynamic range of 60–100 photons per pulse, and stable operation at both 80 K and 100 K. Integration of 60 µm-focal-length microlens arrays boosted the fill factor and spatial resolution, while recent developments include new 3.7 µm-cutoff devices, 2 × 30 and 7 × 8 formats [75], TIA bandwidth enhancements, and impulse responses as fast as 3.5 ns FWHM, demonstrating DRS progress toward high-gain, low-noise, single-photon lidar detectors.

3.2.2. Record-Breaking DRS LM Photon-Counting APDs for Laser Communication

Advancements in the Hg_1−xCd_xTe LMPC e-APDs at Leonardo DRS have significantly improved performance in terms of photon detection efficiency, false-event rate, and timing jitter. These detectors possess single-photon sensitivity, which enables the development of several new LiDAR applications for NASA space missions. Key improvements include the new ROIC blocking design, which reduces glow-induced dark counts, and the successful integration of micro-lens arrays to boost sensitivity [77]. A redesigned light barrier between the ROIC and the absorber effectively reduced the glow-induced dark counts by almost a factor of four (from nearly 35 kHz to 150 kHz), resulting in record-low false event rates (FERs). Without effective blocking layers, the NIR photons emitted by Si transistors can reach the HgCdTe detector, where they can be absorbed and generate photoelectrons. These photoelectrons may then diffuse and experience avalanche gain, leading to an increase in FER. The solution is to use blocking layers to prevent the NIR photons from reaching the absorber. In evaluating the FER at a photon detection efficiency (PDE) of 30%, the median values observed in different 2×8 MWIR e-APD architectures were 3.0 kHz for the 3.7 μm array and 22.6 kHz for the 4.4 μm array with this new design. To enhance detector sensitivity, micro-lens arrays (MLAs) were integrated to focus light onto the p-region, ensuring full avalanche multiplication, and the integration process is shown in a photo of an FPA with alignment features and vias. Gain measurements were conducted on an HDVIP e-APD array with a material cutoff of approximately 4.4 μm, arranged in 4×4 configuration with a 40 μm diode pitch and 80 μm pixel pitch. At a reverse bias of 12.9 V, the gain was 1200, which is typical for MWIR e-APD technology, but at 14.9 V, the gain reached a new record of 6100.

3.2.3. Monte Carlo Modeling of Bandgap-Engineered HgCdTe APDs

The work of Leonardo DRS demonstrates the use of Monte Carlo simulations to model the performance of bandgap-engineered HgCdTe APDs, with device geometries and cadmium compositions ranging from 0.37 to 0.425. The simulations accurately predicted key parameters such as gain, excess noise, and bandwidth by incorporating 3D effects and calibrating with experimental data [78]. For HDVIP complex structures, the non-uniform electric field and spatially dependent impact ionization coefficients lead to varying gains based on where the photoelectrons enter the junction. To model this accurately, electron distributions were calculated using drift-diffusion simulations, which showed higher electron density at the top of the film. For bandgap-engineered films, the non-uniformity is even more pronounced owing to the quasi-electric field created by the bandgap grading. To capture these 3D effects, Monte Carlo simulations seed electrons at multiple radial positions and depths, with a total of 88 sampled positions within the film, as shown in the electron distribution after simulation.

3.2.4. Drift–Diffusion Modeling of Gain and Dark Current in p-Around -n APDs

The simulations showed good agreement with the experimental values from Leonardo DRS at 240 K, particularly the excess noise factor of 1.5 for gains above 100 and the impulse response shape, which matches previous Monte Carlo simulations. The secondary bulge in the impulse response at high gain is attributed to the slower movement of holes, and the wider pulse in the A9144 device (with a 6.5 µm junction) compared with the WA734 device (5.5 µm junction) explains its longer response time [79]. The performance of HDVIP APDs was simulated by varying the gain region widths from 1.5 µm to 7.0 µm. Shorter gain regions (from 7 µm to 2 µm) result in higher gain due to stronger electric fields, but for regions shorter than 2 µm, gain decreases due to ionization dead-space. Shorter gain regions lead to shorter response times. For a 2.5 µm width, the electron response time was 10 ps, and the hole response time was 60 ps, compared to 16.5 ps and 155 ps for a 6.5 µm width. Shorter gain regions exhibit lower excess noise factors. For a mean gain of ≈100, the 6.5 µm region showed higher noise due to a broader gain distribution. While shorter gain regions are generally better, higher fields in these regions lead to increased band-to-band tunneling, which can cause dark current and limit achievable gain. These results emphasize the trade-off between optimizing gain and minimizing dark current in the HDVIP APD designs.

3.3. CEA/Leti HgCdTe APDs for High-Speed and Large-Dynamic-Range LiDAR

3.3.1. Application-Tailored HgCdTe APDs for High-Dynamic-Range LiDAR

CEA/Leti is developing HgCdTe APD detectors for time-resolved and high dynamic range applications, such as free-space optical (FSO) communication and LiDAR. These applications drive the demand for APDs with diameters larger than 10 µm. However, the standard technology at CEA/Leti has been found to create APDs with a significant asymmetry between the central and peripheral regions of the multiplication layer. This asymmetry limits the maximum gain of the APD due to the generation of tunnel current, as well as the achievable bandwidth at high gain. An expanded central multiplication width is required to reduce the tunnel current, while this also results in a lower overall gain. Additionally, the variation in gain across the APD leads to an increase in the excess noise factor. When pushing this technology to large-diameter APDs, gains greater than 100 were achieved, but at the cost of lower central gain. A bandwidth of 10 GHz was observed at unity gain for a 10 µm diameter APD with a large 4 µm multiplication layer at 300 K, demonstrating that carrier collection and extraction in HgCdTe APDs can support 10 GHz operation at room temperature. However, the wide multiplication layer limits the bandwidth at higher gains (M = 40) [80]. For LiDAR applications, a detector module is being developed as part of the H2020 HOLDON project, with the goal of creating a versatile detector suitable for a wide range of atmospheric LiDAR missions. A CMOS ASIC has been designed for hybridization with a large-area HgCdTe APD to meet these requirements. The ASIC employs a capacitive transimpedance amplifier (CTIA) architecture to provide a linear gain of more than six orders of magnitude of the signal. In this setup, the APD gain helps reduce the impact of amplifier noise and enables the use of a larger feedback capacitor, thereby improving the linearity. The high dynamic range is achieved through a variable gain mechanism, offering a range of three orders of magnitude, with a large dynamic range at each gain level. The ASIC also features an on-chip sampling (OCS) mode, which optimizes dynamic range by fast sampling the signal in the ASIC while using a slower ADC for readout, resulting in reduced noise. This configuration provided a dynamic range of 88 dB at the lowest CTIA gain. Initial tests with the HOLDON ASIC hybridized with the HgCdTe APD have validated most of the detector’s functionality, including both continuous and OCS modes, with noise levels close to expectations. The high linear dynamic range was confirmed through measurements showing a linear response over four orders of magnitude and a total dynamic range exceeding six orders of magnitude. The only unconfirmed characteristic thus far is the response time, which is limited to a rise time of 9 ns at the lowest gain, likely due to the APDs’ response time. In the short term, new high-gain APD hybrids will be characterized using the detection chain developed within the HOLDON project, including dedicated control electronics by IDQ. These detectors will be tested in various LiDAR applications by project partners. In the long term, both high-bandwidth optical communication and HOLDON detector applications will benefit from continued advancements in APD technology, aimed at achieving higher gain, reduced excess noise, and enhanced bandwidth.

3.3.2. GHz-Rate Single-Photon Detection APDs for ESA Optical Communications

The application of HgCdTe APD detectors in deep space optical communications was initially demonstrated in the NASA Lunar Laser Communication Demonstration, where a CEA/Leti detector was used to receive 78 Mbps of 16-ary pulse position-modulated (PPM) data at 20 photons per bit at the European Space Agency’s Optical Ground Station (OGS) in Tenerife [81]. This technology has become one of the most promising options for single-photon-level applications, including deep space communications and quantum information processing (Figure 5). In 2021, CEA/Leti presented the architecture and characterization of a four-quadrants detection module, designed for high-data-rate ground-segment detection in deep space optical communications at wavelength of 1.55 μm, in collaboration with the European Space Agency. The detector, featuring a multiplication gain exceeding 150, input-referred noise of 45 electrons RMS at unity gain, and a bandwidth of 450 MHz, enables linear-mode detection of meso-photonic states down to the single-photon level, with high count rates approaching 2 GHz when the signal is distributed across four quadrants [82]. Characterization revealed an excess noise factor of 1.74, detection efficiency of 0.73, and temporal jitter of 160 ps (FWHM) at 10 photons per pulse. The observed high excess noise is attributed to a non-homogeneous gain distribution across the APD area, along with suboptimal optical coupling within the APD, which also contributes to the significant tail in the response time jitter due to the random diffusion of carriers generated outside the multiplication layer. The detector was used to estimate the performance of low-signal free-space optical communication using PPM and on-off keying (OOK) modulation formats. Despite some limitations in the performance of the current detector such as quantum efficiency, excess noise, and temporal jitter, the results demonstrate that a detection probability of 90% can be achieved with a low false alarm rate of 10 photons per bit and a time slot width of 500 ps. For OOK modulation, the bit error rate (BER) was found to be influenced by the quantum efficiency-to-excess-noise ratio (QEFR), with a −3.9 dB penalty at a BER of 2 × 10⁻⁵ for 12.5 photons per bit. The average output voltage of the detector was a function of the incident optical power (and the average number of incident photons, μph) for three different APD bias voltages, corresponding to multiplication gains ranging from 85 to 155. For the dark current module, the total background current measured across the four APDs was 0.59 nA at a bias voltage of −15 V. These promising results were achieved despite detector limitations, and ongoing improvements in APD technology, including the development of larger-area devices with more homogeneous multiplication layers, are expected to reduce excess noise, increase gain, and enhance response time. Additionally, the integration of micro-lenses for better optical coupling will further improve detection efficiency and reduce temporal jitter.

3.3.3. High-Dynamic-Range APDs with Multi-Gain CTIA ROIC for LiDAR Applications

The specific Readout Integrated Circuit was developed for LiDAR applications in 2023 as part of the H2020 project HOLDON. The ROIC uses a CTIA with four different current/voltage conversion gains, owing to capacitors ranging from 10 fF to 10 pF. Additional features such as on-chip sampling, auto-reset, and programmable low-pass filtering were integrated to optimize performance for various measurement chains [25]. The photodiode array was connected via an interconnection circuit to the bonding pads, which were wire-bonded to the ROIC. This allows for flexibility in coupling the ROIC to different APD geometries, depending on the application. The wire-bonded version of the HOLDON detector shows similar performance to the hybrid version, allowing further ROIC development without waiting for a complete CMOS wafer batch. The wire-bonded version was tested with an array of 76 HgCdTe APDs arranged in parallel at 78 K inside a liquid nitrogen-cooled cryostat, each with a 15 μm pixel pitch, forming a 150 μm diameter macro-photodiode. At the highest bias, the APD gain reached 78, with an excess noise ratio of 1.28 and quantum efficiency >70%. At the maximum APD gain, the dark current was approximately 20 pA, aligned with a gain-normalized dark current of 1.6 × 10⁶ e⁻/s. The device demonstrated a high linear dynamic range of more than six orders of magnitude, enabled by the variable CTIA gain. The raw ROIC noise was below the single photon level for the highest ROIC gain and maximum APD gain, indicating excellent sensitivity. The best rise time observed was 6 ns, which meets the requirements for most LiDAR applications where pulse durations are typically ≥10 ns. The relative persistence after a large photonic pulse was 10⁻⁵ at 500 ns and continued to decrease beyond that, with the final value limited by the experimental setup. In the short term, the team plans to test and deliver detector modules with micro-lenses to increase the effective detection area while maintaining a small active APD. These results demonstrate that the wire-bonded version of the detector performs comparably to the hybrid version and is ready for further development. The detector offers excellent sensitivity, high dynamic range, and fast temporal resolution, with improvements expected in future versions.

3.3.4. Four-Quadrant Detector Module for Deep Space Optical Communications

The four-quadrant photon counting detector module with a 200 μm diameter optical collection area per quadrant was developed for deep space optical communications and is scheduled for delivery to the ESA in fall 2022. The detector was characterized to evaluate its performance against the ESA system requirements for deep space optical communications using pulse position modulation (ppm) [83]. The detector meets most of the system requirements, although it has some flaws, including imperfect optical coupling owing to the use of an APD with a small diameter. The high dark count rate requires a high threshold level to meet the required false alarm rate, which reduces the detection efficiency at given signal levels. These issues impact the sensitivity and temporal resolution and could be more problematic when the detector is used for high-order ppm modulation in deep space applications. Despite these flaws, the detector demonstrated the ability to detect 16-order ppm at a slot width of about 0.8 ns, corresponding to 312 Mbps with 90% detection efficiency, a false alarm rate below 10⁻², and less than two photons per bit when all light is concentrated on a single channel. The detector demonstrated single photon detection capability for each of the four channels with a bandwidth close to 400 MHz. For ppm modulation, the detector showed a pulse detection probability of over 90% for a 7-photon signal in an 800 ps time slot, with a false alarm rate below 1%. However, the sensitivity of the full detector was limited by its low quantum efficiency and high dark-count rate. For a 16-ary ppm, this performance corresponds to a data rate of 320 Mbps with less than two photons per bit. The developed detector shows promising performance for deep space optical communications but could benefit from improvements in optical coupling, APD technology, and dark count rate management. With these enhancements, the detector can meet the demanding specifications for ESA deep space applications, including higher data rates and lower photon counts per bit.

3.4. Advancements in Planar APD Research for AIM Applications

3.4.1. HgCdTe e-APD in Short-Wave Infrared Range for Gated Viewing Applications

The SIECK AIM research team presented the properties of SWIR HgCdTe eAPDs for use in gated viewing systems, and gated viewing (GV) enhances image contrast, particularly in cluttered environments (

Table 4. Essential specifications of the ROIC for gated imaging applications [84].

Parameter	Value
Format/pitch	$640 \times$$512/15 \mathsf{\mu }\mathrm{m}$
Input stage	CTIA
Operating modes	GV, GV-CDS, ITR
Charge handling capacity	120,000 e−
Readout noise (w/o CDS)	80 e−
Clock master/PLL	10 MHz/200 MHz
Full frame rate (no CDS)	100 Hz
Full frame rate w/CDS	50 Hz
Subframe steps	1 pixel vertical, 4 pixels horizontal
Scan direction	Programmable independently for rows and columns
Polarity photodiodes	2 ROIC variants for n on p and for p on n

). The number of photoelectrons in the signal depends on factors such as the laser pulse energy, optics focal length and F# number, distance to the object, and reflectance of the object. The eAPDs were made from HgCdTe with a composition of x = 0.4, yielding a cutoff wavelength of 2.55 μm at 150 K. The gain at a fixed reverse voltage depends on the bandgap of HgCdTe material. For these n-on-p devices, a gain of 20 was achieved at a 14 V reverse voltage at temperatures of 140–170 K. The doping profile and width of the space charge region (SCR) need to be optimized to ensure that the gain increases exponentially with reverse voltage, whereas a narrow SCR leads to dark current dominated by tunneling at high reverse voltages. The doping profile of the low-doped n-region in the APD was optimized by adjusting implantation parameters and p-type doping concentration, which increased as the p-type concentration decreased, thereby affecting the electric field distribution under reverse bias. Moreover, the bandgap widening near the surface was optimized to handle the high reverse voltage conditions in avalanche mode. Results of current-voltage (I–V) and gain-voltage plots measured at different temperatures show that for temperatures up to 170 K, the currents are limited by background up to 7 V, with diodes having high p-type concentration exhibiting tunneling current at more negative voltages. Above 170 K, the dark current dominates in both cases, and the slope of the gain–voltage plots decrease as the operating temperature increases [84,85]. FPAs with ROICs featuring fast internal clocks were tested for GV applications. A series of laboratory snapshots were taken at Fraunhofer IOSB in Ettlingen, Germany, using ROIC-V1 to test short integration times and gate control in gated-viewing mode, with the gate length set to 30 ns and shifted in 5 ns steps, which progressively revealed different parts of a tree and house, as shown in a color-coded composite of the full sequence. Ongoing development aims to optimize the doping profile and SCR to achieve consistent exponential gain and better performance at higher temperatures, making these devices viable for longer-distance applications up to 1 km.

For the past decade, the Fraunhofer IOSB has been advancing SWIR laser GV technology, with applications spanning both military and civilian sectors in land and maritime environments [86]. Their work included improving visibility in poor conditions, enabling long-range detection, generating contour images using sliding gates and slope methods, bistatic GV imaging, and imaging through windows (Figure 6). The review of SWIR laser gated-viewing at Fraunhofer IOSB also explores theoretical studies on 3D accuracy and range performance. Future research will focus on exploring the ability of SWIR GV to penetrate pyrotechnic effects, with experimental validation through image simulations.

3.4.2. Extending 3D Range of Laser GV System: Quadratic Model and Longer Laser Pulse Duration

Göhler et al. explored two measures aimed at extending the 3D range of the SWIR laser GV system, building on their previous work in improving 3D reconstruction for dynamic scenes [87]. The SWIR GV system, equipped with a 640 × 512 MCT-based avalanche photodiode focal plane array, had a limited 3D range because of the time difference between two required GV images. To overcome range limitations, researchers enhanced 3D reconstruction by adopting a quadratic model, improving range by up to 83%, and extended laser pulse duration, achieving an additional 21% range increase. In future work, they plan to implement the quadratic model for real-time 3D reconstruction, study the impact of longer laser pulses on the SWIR GV camera capable of correlated double sampling (CDS), and explore pixel-level adjustments in APD gain and CTIA response times to optimize performance. To address the 50 ms delay between slope and plateau images in a 20 Hz system, a method was implemented on an SWIR GV camera with a CDS-enabled ROIC to extract both images from a single laser pulse, enabling 3D imaging of fast-moving objects but initially limiting the range to 3 m [88]. Building on the prior work, Göhler et al. investigated the temporal behavior of detectors in SWIR laser GV system, demonstrating that increasing the APD response time-by reducing the in-pixel CTIA current, tripled the 3D range. They discovered a linear relationship between slope length and signal strength, which was integrated into the 3D reconstruction algorithm to improve range accuracy. Future improvements will focus on refining the model to address overshoot in the quotient profile. Range images using both constant and linear slope models confirmed the linear model ability to provide accurate and signal-responsive range measurements.

3.4.3. Progress in SWIR Long-Range Identification and Smoke Obscuration Penetration

The research at AIM has focused on developing an SWIR module for active imaging, designed to improve long-range reconnaissance capabilities with additional contributions from the Fraunhofer IOSB [89]. This array, fabricated using LPE growth of MCT layers on CdZnTe substrates, features a 2.5 µm cut-off wavelength, which is chosen as the balance between the achievable APD gain and operating temperature (~150 K) for SWIR gated-viewing systems with target distances of approximately 1000 m. A custom ROIC was developed to enable GV functionality, offering a charge handling capacity of 0.12 mio e⁻ and precise control over gate delay. The median noise level was approximately 64 e⁻, operating at average SNR of 245 with M = 1. The system was tested in a field camera demonstrator that integrated an SWIR GV camera, thermal camera, and laser illuminator. The demonstrator demonstrated long-range reconnaissance capabilities, including the ability to see through smoke obscuration. Key performance results include the following: the MCT-based 2D eAPD array achieved a gain of M ≈ 20 at −14 V reverse bias with low excess noise (F ≈ 1.05–1.5). Further tests by Fraunhofer IOSB involved comparing APD gains (M = 20 vs. M = 1) for an improved SNR at a distance of 700 m. In another demonstration, the SWIR GV system was tested for seeing through smoke obscuration by comparing SWIR GV images with visible imaging sensor (VIS) images during pyrotechnic smoke simulation. The results illustrate the effectiveness of SWIR GV in detecting a target at 135 m, even with smoke obscuration, while the VIS sensor struggled. The GV system demonstrated precise time control of delay and integration in 5 ns steps, triggered externally by the laser illuminator. Field trials demonstrated the successful application of the SWIR GV module for identification in degraded visual environments.

3.4.4. Assessment of Gated Viewing at 2.09 µm and 1.57 µm Laser Wavelength

The gated viewing system operating at 2.09 µm laser wavelength in the SWIR range was developed at AIM to assess its potential for security and military application, and this system was compared to a GV system operating at the commonly used 1.57 µm wavelength [90]. The system uses a 640 × 512/15 µm MCT-APD array with sensitivity in the extended SWIR range (0.9–2.55 µm), allowing it to capture both laser wavelengths. The camera offers a FOV of 0.92° × 0.73°, equipped with a 600 mm focal length aspherical F/3 lens. The study confirmed theoretical impacts, such as turbulence-induced scintillation and vegetation reflection, although further investigation is needed to fully explain the observed numerical values. The system successfully demonstrated field measurements at distances of up to 2 km, marking a significant step toward direct comparisons between the two wavelengths. The SWIR GV system was tested on a church tower at a distance of 700 m in order to assess its performance. The APD gain was adjusted to avoid pixel saturation, with M = 6 for λ = 1.57 µm and M = 9 for λ = 2.09 µm. Additionally, laser reflectance measurements were made for human skin and clothing using reference targets at a distance of 70 m. The reflectance values for λ = 2.09 µm were lower for both human skin (~4% vs. ~8%) and clothing (~29% vs. ~40%) compared to λ = 1.57 µm, with decreases of approximately 46% and 20–37%, respectively. The measurements show that the longer laser wavelength (λ = 2.09 µm) results in a significant decrease in reflectance for both tree trunks and leaves compared to λ = 1.57 µm. Specifically, reflectance for the tree trunk dropped from 34.4% to 24.8%, and for the leaf from 20.9% to 10.4%. The trunk-to-leaf contrast ratio increased by 44%, from 1.65 to 2.38, thereby enhancing visibility and contrast. For long-range tests at 2 km, both wavelengths were used to image a tree, with a gate length of 37.5 m and a camera gain of M = 9. The results showed clear visibility of the tree, trunk, and leaves for both wavelengths. The performance of the fabricated APDs relevant to their intended applications is summarized in

Table 5. Comparison of performance among representative material-based APD photodetectors.

Device/Material	Wavelength (nm)	QE (%)	Gain, Voltage (V)
Si APD	400–1100	77	20–400, 150–400 V
InGaAs APD	900–1700	60–70	10–40, 20–30 V
Leonardo-MCT APD	2500–3500	90	66, 14.5 V
DRS- MCT APD	4300	72	6 100, 15.9 V
CEA/Leti- MCT APD	2500–5300	90	2 k SW, 13 k MW
AIM- GV MCT APD	900–2400	75	20, 14 V
PEA2MA3 Pb4I13	250–450	122	0.41 A/W
WSe2 APD	500–800	2000	470, 1.44–3 V

, which provides key parameters among representative material-based APD photodetectors.

3.5. Advancements and Characterization of e-APDs at Shanghai Institute of Technical Physics

Significant progress has been made at the Shanghai Institute of Technical Physics (SITP) in the development of HgCdTe APDs. A variety of single-element and focal plane array devices covering the SW, MW, and LW infrared ranges have been successfully fabricated. SITP researchers have concentrated on evaluating the key performance metrics of these devices, including the gain, dark current, and excess noise. Recent innovations in SW and MW e-APDs represent a major leap forward in infrared detection technology, further pushing the boundaries of infrared photodetection technology. For single-element SW-HgCdTe APDs with a 2.57 µm cutoff, a gain of approximately 100 was achieved at reverse bias of 25 V, with I_D of 1.47 × 10⁷ A/cm² at 130 K. In MW-HgCdTe APDs, increasing the p-region doping concentration reduced the overall dark current density and prevented the sudden rise in dark current at high bias and temperature, and a lower Cd composition was proposed as a strategy for suppressing GNDCD. A 50 µm pitch 128 × 128 array of HgCdTe APDs with a 4.88 µm cutoff wavelength (x_cd = 0.307) exhibited I_D of less than 1 × 10⁷ A/cm² at an 8 V reverse bias, with a gain exceeding 1000 at an 11 V reverse bias. Another 128 × 128 array with a composition of x_cd = 0.29 reached a gain of 1570 at a 9.8 V reverse bias, with an average excess noise factor of 1.25 at a gain of 133, and a noise equivalent photon count of approximately 12 at a gain of 113. By reducing the absorption region thickness, the response bandwidth of Hg_0.79Cd_0.31Te APDs reached 635 MHz under a 1 V reverse bias. Additionally, the imaging capabilities of a 320 × 256 MW FPA were demonstrated, confirming the low noise, high sensitivity, and fast imaging performance of HgCdTe APDs under linear avalanche gain conditions [56,91,92,93].

Through optimization of the multiplication region and annealing process, high-definition infrared imaging with reduced noise and tunneling current was achieved in 640 × 512 array APDs, highlighting the significance of implanted area size for enhanced imaging performance, as explored by Chen et al. [94]. This study presents the effects of optimizing the multiplication region in HgCdTe APDs with a 10 μm implanted area diameter. As the implanted area decreased, the multiplication region narrowed, leading to an increased BBT current and a higher electric field. To address this, refining the annealing process reduces the BBT current and improves imaging performance. Figure 7 shows that optimized devices with larger multiplication regions resulted in fewer bad pixels, improved infrared imaging at short integration times, and accurate identification of concealed objects. These results demonstrate that HgCdTe APDs with optimized gains are suitable for rapid, high-quality infrared imaging.

Li et al. optimized the avalanche structure geometry and doping parameters in MWIR APDs, achieving a high gain of 1876 (6153) and a dark current as low as 10⁻¹⁰ (10⁻⁹) A at −10 (−10.5) V [55]. Combined with the blackbody detectivity, the peak photodetectivity D* is obtained as 2 × 10¹⁴ cm Hz^1/2 W⁻¹ at the bias voltage −7.1 V. The proposed theoretical model reveals that both the ionization avalanche and BBT effects contribute to dark current, but the ionization avalanche effect is more significant, challenging the conventional belief that the BBT dominates at high bias. The findings of this study provide a valuable framework for the performance evaluation of MCT APDs and offer guidance for structural design and optimization.

He et al. investigated the temperature-dependent I–V characteristics of planar (n⁺/n⁻/p) LWIR HgCdTe APDs with cutoff wavelengths (λc) of 11.5 μm at 80 K. The impact of BBT on F measurements was explored through noise power spectral density (PSD) and noise figure meter analyses. A low F value 1–1.27 was achieved at a gain (M) of 6 at −2 V, and it was extended to a higher gain (M = 23) at −3 V. The study also analyzed bias-dependent dark noise, proposing that the BBT component may partially experience avalanche multiplication in the LWIR devices. Reducing BBT current through an appropriate annealing process could be an effective way to reduce the dark current and noise [95].

Lin et al. explored the frequency response of MWIR HgCdTe APDs, which are essential for free-space communication and spectroscopy, in addition to the commonly studied DC characteristics. The 20 μm-diameter device demonstrates a low dark current (~10⁻¹³ A at low reverse bias), high gain (1356 at 11.4 V), low excess noise (1.3 at 78 K), and a responsivity of 1.48 A/W at 4.5 μm under 0 V. The response bandwidth reached 635 MHz at 1 V. The proposed upside-down structure suggests achieving over 30 GHz with a 1 μm intrinsic region, and it can be concluded that the vertical quasi-electric field distribution induced by the top electrode contacts enhances the device bandwidth. The optimized structure positioned the p-electrode above the absorption region, enabling photon-generated carriers to enter the multiplication region vertically [96].

The recent work by Zhu et al. introduces a modified fully depleted absorption multiplication (MFDAM) Hg_1−xCd_xTe MWIR APD—using band-gap engineering and composition grading—to achieve unprecedented performance at 160 K, including a multiplication gain of 189 with excess noise <1.4, a noise equivalent power <7.2 × 10⁻¹⁶ W/√Hz at 3.5 µm, and a gain-normalized dark current density consistently below 2 × 10⁻⁶ A/cm² (Figure 8), demonstrating signal detection beyond 200 km, establishing these devices as leading candidates for high-temperature MWIR and photon-starved applications [97].

3.6. Research on HgCdTe APD Focal Plane Technology at Kunming Institute of Physics

The Kunming Institute of Physics (KIP) has conducted extensive research on HgCdTe APD focal plane technology, focusing on enhancing performance metrics such as gain, noise, and response time. In 2022, KIP achieved a significant milestone by developing 320 × 256 electron-multiplying HgCdTe MW linear avalanche infrared focal plane detector. The findings indicate that the average gain of the HgCdTe APD focal plane chip is 166.8, with a gain non-uniformity of 3.33% at reverse bias of 8.5 V, as shown in Figure 9. The gain-normalized dark current for the APD device ranges from 9.0 × 10⁻¹⁴ A to 1.6 × 10⁻¹³ A under a reverse bias of 0 to −8.5 V, while the noise factor F is observed to be between 1.0 and 1.5. Additionally, imaging demonstrations with the HgCdTe APD focal plane revealed a promising imaging performance, demonstrating the potential of this technology for infrared applications [98,99]. These advancements highlight the detector potential for high-sensitivity infrared applications and capability to operate effectively in demanding conditions.

4. Conclusions and Perspectives

In this paper, we present extremely low dark counts at leading research institutions by redesigning the various process links, then directly measure gain and excess noise factors using focal plane array data and modeling and evaluate additional metrics such as false-event rate, photon-detection efficiency, and timing jitter on larger-format arrays. Compared to other infrared APD materials, HgCdTe detectors deliver broadband coverage from SWIR to MWIR, near-unity quantum efficiency, high avalanche gain with minimal excess noise, low dark current, and rapid response, making them ideally suited for demanding applications like LiDAR, deep space exploration, and atmospheric sensing. For further enhance performance, grading the composition in the p-type absorption region can create a quasi-electric field that reduces carrier diffusion and preserves responsivity, and future efforts will prioritize enhancing quantum efficiency and optimizing minority carrier lifetimes. Guided by theoretical insights, innovative material processes and APD architectures are driving these devices toward superior performance, while integrating tunable bandgap materials with application-specific circuit designs offers a pathway to detectors with enhanced sensitivity and reduced excess noise.