In-sensor computing, which integrates signal acquisition and information processing within image sensors, is emerging as a promising route toward low-power artificial intelligence hardware. However, its realization has been hindered by the lack of scalable pixel architectures capable of combining high-quality photodetection with computing functionality. In this work, we propose an in-sensor convolution scheme enabled by a thin-film pinned photodiode (TF-PPD) structure. The TF-PPD employs a photogate (PG) that enables dynamic tuning of pixel sensitivity, thereby facilitating the direct implementation of convolution operations within the sensor array. The convolutional neural network (CNN) implemented with the proposed TF-PPD-based convolution technique was experimentally validated through image demonstrations. The proposed technique achieved 98.6% accuracy on the modified National Institute of Standards and Technology (MNIST) and 94.9% on captured handwritten digits, while reducing the computational cost by 39.7%. These results demonstrate that the TF-PPD provides both stable photodetection and embedded computing capability, highlighting its potential as a scalable device-level solution for next-generation intelligent vision sensors.

This letter presents a single-exposure high dynamic range (HDR) image sensor based on a thin-film pinned photodiode (TF-PPD) pixel structure. The proposed technique efficiently exploits the unique structural features of the TF-PPD architecture to achieve high dynamic range operation while retaining the inherent advantages of thin-film photodiodes, such as wavelength extensibility and ease of integration. Specifically, the photodiode (PD) under the photogate (PG) is utilized as a large photodiode (LPD), while the PD in the floating diffusion (FD) region is employed as a small photodiode (SPD). This enables single-exposure HDR imaging without requiring additional circuitry or complex fabrication steps. Furthermore, by adjusting the integration-time ratio between the LPD and SPD, an additional effective dynamic range can be achieved. The prototype image sensor demonstrates an effective dynamic range of 100 dB with an integration-time ratio of 5:1 between the LPD and SPD. The proposed technique successfully reproduces both bright and dark regions under single-exposure conditions, confirming its capability for efficient HDR imaging within a compact pixel design.

Heavy-metal-free III-V semiconductor-based colloidal quantum dots (CQDs), such as InAs, are promising candidates for near- and short-wave infrared detection. However, up-to-date research efforts remain mainly limited to wavelengths below 1100 nm due to challenges in synthesis, junction formation, and passivation for large diameter InAs quantum dots. Systematic investigations into device design, reverse dark current mechanisms, and trap distributions in larger InAs quantum dots remain limited. Here, we report a thin-film PIN heterojunction colloidal InAs (1200 nm) photodiode stack with amorphous indium gallium zinc oxide and copper(I) iodide transport layers. To the best of our knowledge, the device exhibits one of the lowest reported dark current densities of 4.7 μA/cm2 at -1 V and 298 K, which decreases to 3.6 nA/cm2 at 220 K. Temperature-dependent current-voltage characteristics and activation energy analysis confirm thermally driven dark current increasing with applied field. Impedance spectroscopy reveals the dominant deep trap states within the InAs CQD layer, being tail states of the conduction band that reach down to ∼0.4 eV below the band edge, with a density of ∼2 × 1016 cm-3. The temperature-induced increase in carrier density and reduction in built-in potential within the depleted InAs layer reflect trap filling and Fermi level pinning in the N and P layers. The trapping-detrapping induced noise reduces the specific detectivity (D*) at -1 V by 1.97 orders at 1 Hz and by 1.52 orders of magnitude at 10 Hz relative to the shot-noise-limited baseline. At frequencies ∼ ≥500 Hz the D* approaches the calculated limit of 2.5 × 1011 Jones. Finally, we demonstrate infrared imaging by monolithically integrating the photodiode with a Si read-out IC, enabling imaging beyond the spectral range of CMOS sensors.

Photodetector technologies based on emerging semiconductors—for example, organic semiconductors, halide perovskites, quantum dots, low-dimensional semiconductors and metal oxides—hold considerable promise for next-generation optoelectronics. However, the breadth and multidisciplinarity of this field, alongside its diverse range of applications, have resulted in inconsistent performance characterization and reporting practices, hindering the effective benchmarking of these technologies. Here we present a consensus among researchers from academia and industry on accurately capturing the key performance metrics of photodetectors based on emerging semiconductors and utilizing the photoelectric effect. We analyse their underlying assumptions, discuss common misunderstandings, and provide guidelines for accurate characterization and reporting. Additionally, we discuss the benchmarking of these photodetector technologies with respect to diverse applications. We expect that these comprehensive guidelines for characterization, reporting and benchmarking will accelerate and streamline further advancements in the field, propelling emerging photodetector technologies towards their full potential. This Consensus Statement discusses common misunderstandings about photodetector performance characterization and reporting, and offers recommendations for standardized practices.

Colloidal quantum dots are increasingly employed in display systems for their unbeatable light output quality. At the same time, their light absorbing qualities are highly appreciated in photodetector and sensor systems, especially in the infrared wavelength spectrum beyond the silicon cut-off at 1100 nm. In this paper, we introduce Pb-free absorbers in thin-film photodetector stacks with external quantum efficiencies above 30% in the short-wave infrared (SWIR) range. These devices are integrated in high-resolution and high pixel density image sensors enabling augmented vision.

This work presents a monolithically integrated short-wavelength infrared (SWIR) image sensor based on indium arsenide (InAs) quantum dot photodiodes (QDPDs). The thin-film photodiode (TFPD) architecture enables direct integration on silicon readout integrated circuits (ROICs), eliminating wafer-to-wafer bonding and providing a scalable, RoHS-compliant alternative to lead-based colloidal quantum dot (CQD) devices. The proposed 3T pixel design incorporates dual conversion gain (DCG), enabling wide dynamic range imaging. The fabricated prototype achieves external quantum efficiencies of 28% at 1200 nm and 4.8% at 1400 nm, together with a dynamic range of 83.5 dB. A frame-based digital correlated double sampling (CDS) scheme stores the reset level in the digital domain and subtracts it after integration, thereby suppressing reset kTC noise and mitigating random telegraph signal (RTS) noise. Imaging demonstrations highlight SWIR-specific functionalities, including material discrimination, imaging through smoke, and transmission through silicon wafers. A performance comparison with previously reported SWIR pixels further confirms the competitiveness of the proposed InAs QDPD imager. These results establish InAs QDPDs as a promising platform for next-generation SWIR imaging, combining high sensitivity, extended spectral coverage, and scalable integration.

In this article, the X-ray radiation effects on colloidal quantum dot photodiode (QDPD)-based short-wave infrared (SWIR) complementary metal-oxide semiconductor image sensors (QD-CISs) are studied. Individual QDPD, silicon readout IC (Si-ROIC), and QD-CIS are evaluated together for a comprehensive analysis. The dark current, activation energy, and external quantum efficiency (EQE) of samples are investigated before and after irradiating with 58.2 keV of X-ray radiation, which has a different total ionizing dose (TID) range from 22 to 220 krad. X-ray irradiation on Si-ROIC induces mid-band gap trap states and increases the dark current according to the increasing TID. However, for the QDPD, despite an increase in the TID, the dark current reduces and the EQE slightly enhances at the SWIR wavelength. The QD-CIS shows a decrease in the dark current like the QDPD results, as the TID increases. The activation energy of QD-CIS rarely changes regardless of TID amounts. The X-ray radiation effect on QDPD results in enhanced performance, and this effect continues in the integrated QD-CIS, whereas the effect of Si-ROIC degradation is minor in the current experimental range. Thus, these findings provide significant insights into the utilization of QD-CIS in various X-ray applications.

Colloidal quantum dot sensors are disrupting imaging beyond the spectral limits of silicon. In this paper, we present imagers based on InAs QDs as alternative for 1st generation Pb-based stacks. New synthesis method yields 9 nm QDs optimized for 1400 nm and solution-phase ligand exchange results in uniform 1-step coating. Initial EQE is 17.4% at 1390 nm on glass and 5.8% EQE on silicon (detectivity of $7.4\times 10^{9}$ Jones). Using metal-oxide transport layers and>300 hour air-stability enable compatibility with fab manufacturing. These results are a starting point towards the 2nd generation quantum dot SWIR imagers.

In this paper, we present thin-film photodetector (TFPD) image sensors for the short-wave infrared (SWIR) range. Monolithic integration of quantum dot (QD) absorbers enables quantum efficiency of 70% at 1400 nm and pixel pitch below 2 μm. We present image sensors on custom CMOS readout fabricated using 130 nm node. We review latest advancements on the photodiode stack and the pixel engine, including the thin-film pinned photodiode architecture. Furthermore, we study the manufacturing flows to realize full wafer capability for volume processing. QD image sensors are paving the way to add augmented vision into future XR systems with extra functionalities.

Colloidal quantum dots (CQDs) have emerged as promising materials for thin film photodiodes (TFPDs) in the short-wavelength infrared detection range, offering an alternative to III–V and HgCdTe-based TFPDs. However, optimizing the structure of CQD-based TFPDs remains a challenge, as it involves a delicate balance between reducing dark currents and enhancing carrier extraction efficiency. In this study, we explore the influence of varying the thickness of CQD layers to achieve a highly efficient photodiode. Our investigations reveal a continuous reduction in the dark current as the CQD layers become thicker, but we observe fluctuation in the external quantum efficiency (EQE). To shed light on this relationship between dark current density (Jdark) and EQE, we conduct capacitance measurements and employ optical simulations. From the capacitance measurements, they demonstrate an increased depletion width with varying CQD thickness, apart from layers exceeding 500 nm in thickness. Leveraging optical simulations, we propose an optimal thickness for CQD-based TFPDs and compare its EQE performance. The optimized CQD-based TFPD exhibits a Jdark of 4.1 μA/cm2 and EQE of 56.5%, and the highest specific detectivity, based on the assumption of shot noise dominance, is 1.78 × 1012 Jones at a wavelength of 1420 nm.

Colloidal quantum dots (CQDs) are cutting-edge optoelectronic semiconductor nanocrystals that enable short-wave infrared (SWIR) vision by a widely tunable SWIR light absorption. Thanks to the advances in CQD surface ligand engineering, SWIR detectors and emitters will soon find their way into products. The CQD-based optoelectronic devices are being optimized by adapting the size of CQDs and selection of the ligands, and yet, the measurement schemes of energy band structure based on different ligands and processes of ligand exchange are not systematically studied. In this work, we systematically characterize the energy band structure of PbS (absorbing at different SWIR wavelengths) and InAs with various ligands for both solid-state and liquid-phase ligand exchange (LPLE) processes [solid-state ligand exchange (SSLE) and LPLE] by using ultraviolet photoelectron spectroscopy. The deduced energy band structures reveal that the apparent energy difference between the Fermi and valence band maximum, |E F - E VBM|, largely depends on the physical density and distribution of the CQDs within the probing area. Transmission electron microscopy images, X-ray photoelectron spectroscopy, atomic force microscopy, and variable angle spectroscopic ellipsometry reveal details of the CQD distribution, surface elemental profile, and topologies and how they affect the observed energy band structure. We demonstrate that the multistep coating improves the CQD distribution and packing density, resulting in more reliable and reproducible results that represent the bulk CQD film energy band structure. The comparison of solid and liquid phase ligand-exchanged PbS and InAs SWIR CDQs energetics indicates that the LPLE ensures more uniform dispersion and a high packing density of CQDs regardless of the solution concentration. The photoemission-deduced energy band structures are validated by fabricating thin-film photodiodes using SWIR SSLE PbS and LPLE In(As,P) CQDs. The Fermi-referenced band structures of the fabricated full photodiode stacks including band offsets and bending are discussed to improve our understanding of the device working principles and to further optimize the devices.

Image sensors are must-have components of most consumer electronics devices. They enable portable camera systems, which find their way into billions of devices annually. Such high volumes are possible thanks to the complementary metal-oxide semiconductor (CMOS) platform, leveraging wafer-scale manufacturing. Silicon photodiodes, at the core of CMOS image sensors, are perfectly suited to replicate human vision. Thin-film absorbers are an alternative family of photoactive materials, distinguished by the layer thickness comparable with or smaller than the wavelength of interest. They allow design of imagers with functionalities beyond Si-based sensors, such as transparency or detectivity at wavelengths above Si cutoff (e.g., short-wave infrared). Thin-film image sensors are an emerging device category. While intensive research is ongoing to achieve sufficient performance of thin-film photodetectors, to our best knowledge, there have been few complete studies on their integration into advanced systems. In this paper, we will describe several types of image sensors being developed at imec, based on organic, quantum dot, and perovskite photodiode and show their figures of merit. We also discuss the methodology for selecting the most appropriate sensor architecture (integration with thin-film transistor or CMOS). Application examples based on imec proof-of-concept sensors are demonstrated to showcase emerging use cases.

We report a high-speed low dark current near-infrared (NIR) organic photodetector (OPD) on a silicon substrate with amorphous indium gallium zinc oxide (a-IGZO) as the electron transport layer (ETL). In-depth understanding of the origin of dark current is obtained using an elaborate set of characterization techniques, including temperature-dependent current-voltage measurements, current-based deep-level transient spectroscopy (Q-DLTS), and transient photovoltage decay measurements. These characterization results are complemented by energy band structures deduced from ultraviolet photoelectron spectroscopy. The presence of trap states and a strong dependency of activation energy on the applied reverse bias voltage point to a dark current mechanism based on trap-assisted field-enhanced thermal emission (Poole-Frenkel emission). We significantly reduce this emission by introducing a thin interfacial layer between the donor: acceptor blend and the a-IGZO ETL and obtain a dark current as low as 125 pA/cm2 at an applied reverse bias of -1 V. Thanks to the use of high-mobility metal-oxide transport layers, a fast photo response time of 639 ns (rise) and 1497 ns (fall) is achieved, which, to the best of our knowledge, is among the fastest reported for NIR OPDs. Finally, we present an imager integrating the NIR OPD on a complementary metal oxide semiconductor read-out circuit, demonstrating the significance of the improved dark current characteristics in capturing high-quality sample images with this technology.