Benefiting from the inherent capacity for detecting longer wavelengths inaccessible to human eyes, infrared photodetectors have found numerous applications in both military and daily life, such as individual combat weapons, automatic driving sensors and night-vision devices. However, the imperfect material growth and incomplete device manufacturing impose an inevitable restriction on the further improvement of infrared photodetectors. The advent of artificial microstructures, especially metasurfaces, featuring with strong light field enhancement and multifunctional properties in manipulating the light–matter interactions on subwavelength scale, have promised great potential in overcoming the bottlenecks faced by conventional infrared detectors. Additionally, metasurfaces exhibit versatile and flexible integration with existing detection semiconductors. In this paper, we start with a review of conventionally bulky and recently emerging two-dimensional material-based infrared photodetectors, i.e., InGaAs, HgCdTe, graphene, transition metal dichalcogenides and black phosphorus devices. As to the challenges the detectors are facing, we further discuss the recent progress on the metasurfaces integrated on the photodetectors and demonstrate their role in improving device performance. All information provided in this paper aims to open a new way to boost high-performance infrared photodetectors.

High-stability photovoltaic devices are crucial for low-power or passive applications in fields such as renewable energy, wearable electronics, and deep-space exploration. However, achieving stable and controllable doping in two-dimensional (2D) materials remains challenging, hindering the optimization of photovoltaic performance. Here, we fabricate three high-performance, self-driven photodetectors based on layered WSe₂ with varying doping concentrations. By leveraging asymmetric Schottky barriers and introducing a defect-free, high-bandgap intrinsic region with a long mean free path, we construct a positive–intrinsic–negative (PIN) vertical homojunction that significantly enhances the photogenerated voltage, photon absorption, and carrier transport efficiency. The resulting PIN junction exhibits a photogenerated voltage of up to 0.58 V, a responsivity of 0.35 A/W, and an external quantum efficiency of 83.9%. Moreover, it maintains a reverse saturation current as low as 0.2 nA at 430 K. These results provide a promising route toward the development of high-responsivity, high-stability van der Waals devices and highlight the potential for 2D material-based technologies to operate reliably under extreme conditions.

Miniaturized infrared spectrometers are vital for portable sensing, yet most single-detector computational schemes prioritize spectral encoding while overlooking weak-signal requirements in mid-infrared detection, where high absorption, low dark current, and internal gain are essential. Here, we propose and numerically validate a voltage-programmable spectral sensing concept based on a graded-bandgap HgCdTe avalanche photodiode. In this design, the composition gradient is repurposed as an active spectral-encoding coordinate. Through bias-dependent depletion, the device selectively accesses different absorption depths, while an N/P double-buffer architecture confines the high electric field within the wide-bandgap multiplication region, preserving avalanche gain while suppressing tunneling-related dark current in the absorber. Crucially, the multi-physics TCAD framework—incorporating wavelength-dependent absorption, carrier transport, tunneling, recombination, and impact ionization—is rigorously calibrated against experimental data from a baseline HgCdTe diode. These experimentally anchored simulations reveal distinct electrical fingerprints across 3–7 μm, driven by the coupled effects of absorption depth, photon flux, and wavelength-dependent avalanche path length. Compared to a conventional PIN structure, this architecture reduces dark-current density by one to two orders of magnitude at comparable gain. By processing these fingerprints with a noise-aware dual-branch neural network, discrete-band spectral reconstruction achieves a mean R2 of 0.9894 and a total-energy error of 3.21%. This work establishes a detector-physics-driven route toward filter-free, gain-assisted mid-infrared spectral sensing.