Photodetectors (PDs) can convert a light signal into an electrical signal, typically in the form of voltage or current. Flexible PDs with weak-light detection capabilities play a crucial role in wearable health monitoring^[1], imaging^[2], environmental monitoring^[3], and other fields^[4]. However, weak signals are frequently disturbed or even drowned by noise, which typically stems from background light, dark noise of PDs and noise of amplifier circuits, making it difficult to detect weak light. For decades, researchers have been dedicated to develop weak-light detectors, with silicon photomultiplier tubes and avalanche diodes serving as mainstream technologies. Relying on high-performance materials such as Si, Ge, GaAs, and LnP^[5], these detectors exhibit advantages like high sensitivity, excellent stability, and low noise levels^{[6, 7]}. Nevertheless, these detectors still have design limitations. For example, to achieve significant photoresponse, thicker material layers are often required, which not only increases the brittleness of the devices but also drives up costs and imposes stricter requirements on manufacturing processes, thereby limiting applications in the field of flexible electronics. Therefore, there is an urgent need to select suitable light absorbers and design suitable device structures to extract useful optical signals from noisy signals to solve the above problems.

In light of this, there is an urgent need to select ideal light-absorbing materials, coupled with well-designed device structures that can effectively identify and extract weak-light signals while suppressing noise interference. Since 2013, under the guidance of Professor Henry J. Snaith^[8−10], perovskite materials have gradually gained the attention of researchers and have been extensively studied. The popularity of perovskite materials stems from their remarkable optoelectronic properties, including tunable bandgaps^[11], high light absorption coefficients (10⁵ cm⁻¹), superior photoluminescence quantum yields^[12], long diffusion lengths^[13], as well as simplified fabrication processes and relatively low costs^[14]. Generally speaking, perovskite materials have a structure similar to ABX₃, where the A-site typically represents FA⁺, MA⁺, or Cs⁺, the B-site is Pb²⁺ or substitution using homovalent (Sn²⁺/Ge²⁺)/heterovalent (Cu⁺/Ag⁺/Bi³⁺/Sb³⁺) elements^[15−17], and the X-site represents the halide elements Cl⁻, Br⁻, or I^{−[18, 19]}. By precisely adjusting the ratio of halide elements, the bandgap of the material can be flexibly adjusted, and the optoelectronic properties can be modified or changed by B-site substitution.

Given the diverse preparation strategies for perovskite materials, they can seamlessly integrate into flexible and stretchable substrates, effectively compensating for the limitations of traditional rigid electronic devices in specific functions and filling the gaps. To meet the individual requirements of different application scenarios, we can flexibly design a multitude of PDs, encompassing highly biomimetic optoelectronic devices and innovative wearable medical equipment. As modern technology advances rapidly, these soft yet highly efficient photoelectric sensors demonstrate remarkable potential in fields such as prosthetic technology and human-computer interaction, ushering in boundless possibilities for future intelligent living and healthcare.

In light of this, this review systematically overviews the latest research in perovskite photodetection technology (Fig. 1). Initially, we provide a concise overview of the chemical and physical properties of perovskite materials, analyzing their performance in weak-light detection. Simultaneously, we delve into the fundamental principles of the photoelectric effect, laying a solid theoretical foundation for subsequent discussions. Subsequently, we focus on the fabrication techniques of perovskite materials, thoroughly examining the advantages and application scenarios of multiple preparation methods. Additionally, we thoroughly analyze the key strategies for enhancing the performance of perovskite weak-light PDs from the perspectives of performance improvement (including material optimization and device structure design) and light absorption enhancement. On this basis, we further envision the vast application prospects of perovskite weak-light PDs, ranging from wearable health monitoring to smart sensing technology, all showcasing their immense potential. Finally, we briefly outlook the challenges faced by perovskite weak-light PDs in their commercialization process and propose a series of targeted development strategies aimed at improving their long-term stability and safety.

With its excellent light absorption properties, high carrier mobility and tunable bandgap structure, perovskites show great potential for application in the field of PDs. The working principle of perovskite PDs is rooted in the photoelectric effect, where photon energy absorbed at the material’s surface, generating electron−hole pairs. Subsequently, under the action of an electric field, the electrons and holes migrate towards opposite electrodes and are collected, generating a photocurrent in the process. This sequence transforms optical signals into electrical signals, enabling the detection of optical signals.

To translate these theoretical advantages into high-performance devices, the selection and innovation of preparation methods are critical. By precisely adjusting precursor ratios, optimizing synthesis process parameters, and exploring novel preparation techniques, high performance PDs can be acquired, laying a solid foundation for the further advancement of PDs.

PDs convert incident optical signals into electrical signals. Based on their operating principles, they can be generally categorized into photoconductive^[29] and photovoltaic^[30−32] detectors. Photoconductive detectors are devices manufactured utilizing the photoelectric effect in semiconductors, with photoconductors (Fig. 2(b)) being a typical example. Their operation relies on a constant bias voltage, ingeniously transforming changes in resistance induced by light into current variations, exhibiting a nearly linear volt-ampere characteristic. In contrast, photovoltaic detectors employ the photovoltaic effect, where a P−N junction (Fig. 2(a)) naturally induces a photogenerated potential difference across its terminals when illuminated, thus producing the photovoltaic effect. These detectors are renowned for their low dark current and high detectivity, although they are slightly inferior in internal gain and responsivity compared to photoconductive detectors.

From a structural perspective, PDs can be further divided into two-terminal and three-terminal configurations. Among two-terminal structures, based on electrode layout, they can be subclassified into vertical and horizontal types. The vertical type (Fig. 2(a)) boasts a compact design with each layer’s thickness precisely controlled within a few hundred nanometers. The short distance between photoexcited electron−hole pairs and electrodes, coupled with the strong driving force of internal electric fields, contribute to their exceptional response speed. In contrast, horizontal two-terminal devices (Fig. 2(b)) feature relatively distant electrodes, often requiring an external voltage to facilitate effective carrier separation and collection. Regarding three-terminal optoelectronic devices, a gate is introduced to the traditional two-terminal structure (Fig. 2(c)). This design enables precise control over the transport and recombination processes of photogenerated carriers by adjusting the gate voltage, mimicking the efficient electron−hole pair separation mechanism found in photodiodes. Furthermore, the intervention of the gate voltage modulates the induced photocurrent, granting these detectors unprecedented sensitivity beyond conventional photodiodes.

Among the numerous preparation strategies, the fabrication processes of perovskites can be broadly classified into two categories: dry and wet methods. Dry preparation methods focus on direct reactions in the gas phase or solid state, exemplified by vacuum deposition^[23] and chemical vapor deposition (CVD)^{[21, 33, 34]}. These methods exhibit significant potential for industrial applications, characterized by their high efficiency, high quality, and suitability for large-scale production. In contrast, wet preparation methods rely on chemical reactions and molecular assembly in a liquid environment, including techniques such as drop casting^[35−37], solution spin-coating^{[38, 39]}, spray coating^{[40, 41]}, or inkjet printing^[42−44]. With their advantages of simplicity, cost controllability, and tunable composition, wet preparation methods have become the preferred choice for materials fabrication.

Dry preparation primarily encompasses vacuum evaporation and CVD, with the preparation schematic illustrated in Figs. 3(a)−3(c). The vacuum evaporation process (Fig. 3(a)) operates as follows: A controlled vacuum is maintained within the chamber, and specific inert gases are introduced. By heating the evaporation source, the required organic and inorganic materials for the perovskite layer are deposited onto the substrate surface, forming a dense film. Due to the uniform atmosphere within the chamber, this process exhibits high controllability, excellent film deposition uniformity, and high experimental reproducibility. However, its limitations lie in the relatively slow preparation rate and the high-cost investment. CVD, on the other hand, is the most commonly used preparation method for inorganic perovskite growth. As shown in Fig. 3(b), the preparation process involves the diffusion of reactant gases towards the substrate surface, where they are adsorbed and undergo chemical reactions to form solid deposits^[21]. The resulting gaseous by-products then detach from the substrate surface and are expelled. Nevertheless, during the heating process, the organic components within organic−inorganic hybrid materials tend to decompose or dissociate below the evaporation temperature required for the inorganic components. This renders certain limitations when using single-source evaporation deposition techniques for organic−inorganic hybrid perovskites. To address this, an improved source vapor delivery system can be employed, utilizing a two-tube CVD (2T-CVD) method instead of the conventional one-tube CVD^[33]. As shown in Fig. 3(c), this involves using two separate delivery systems for the two precursors, preventing them from mixing before reaching the vicinity of the substrate surface. Research findings indicate that the utilization of the improved 2T-CVD method enables the growth of smoother and denser films. With the development of CVD, by controlling the growth conditions, high-quality perovskite materials with various morphologies such as micro-sheets^[48], micro/nanowires^[49], and nanospheres^[50] can be further grown.

The wet preparation of perovskites involves initially formulating the required precursor solutions, followed by techniques such as drop-casting, solution spin-coating, spray-coating, or inkjet printing, with subsequent annealing and crystallization to obtain the corresponding perovskite materials. The preparation schematic is illustrated in Figs. 3(d)−3(g). Being a relatively simple preparation method, drop-casting only requires the precursor solution to be dropped onto the substrate, leveraging the spontaneous diffusion of the solution to achieve the desired material (Fig. 3(d))^[45]. However, due to its poor uniformity and reproducibility, drop-casting faces certain limitations in practical applications. Spin-coating method is a more refined preparation technique as shown in Fig. 3(e). The precursor solution is dispensed onto the center of a substrate fixed on a spin coater. Then, the spin coater is activated to rotate the substrate, spreading the solution evenly under centrifugal force to form a thin film. The growth process of perovskite crystals encompasses three main stages: solution supersaturation, nucleation, and crystal growth. When the precursor solution is spread on the substrate, the solvent rapidly evaporates, leading to an increase in solute concentration. As the solvent continues to evaporate, the solution gradually reaches a supersaturated state. When the Gibbs free energy exceeds the nucleation energy barrier, new nuclei begin to form. Subsequently, these nuclei gradually grow and interconnect to form a continuous perovskite film. Spray-coating involves atomizing the dispersion liquid through a spray gun, controlling the spray pressure, speed, and angle to achieve uniform coating thickness and surface quality, and the process is shown in Fig. 3(f). Its advantages lie in its simplicity of operation, uniformity, and high-quality coating, making it suitable for various materials and large-scale production. Nevertheless, issues such as paint waste, environmental pollution, and difficulty in nozzle maintenance may arise during the spraying process. Inkjet printing, as an advanced printing technology, enables non-contact, maskless, and high-quality patterned deposition, as shown in Fig. 3(g). It is compatible with diverse materials like paper, fabric, and plastic, catering to various industry needs. However, inkjet printing suffers from relatively slow printing speeds, limited batch printing capabilities, and small ink cartridge capacities, which restrict its application range to some extent.

Benefiting from the diverse preparation strategies and flexible process controls of perovskites, they can be perfectly integrated with flexible substrates. These diverse preparation methods encompass a wide range of techniques, including vapor deposition, drop-casting, spin-coating, and spray-coating, each of which effectively enhances the film quality and thickness uniformity of the materials, achieving varying degrees of optimization. By precisely controlling the temperature during the preparation process and the post-treatment process, researchers can accurately manipulate the surface morphology of perovskite films, obtaining materials that not only maintain high photoelectric conversion efficiency but also exhibit excellent flexibility and mechanical stability. With these advantages, perovskites have shown great potential for applications in wearable health monitoring, smart sensing and other fields.

The weak-light detection ability of PDs directly affects the accuracy and reliability of their applications. In this chapter, the principle that PDs can maintain high detection accuracy in weak-light conditions is understood through various parameters. Subsequently, we provide an overview of the methods commonly used to enhance the weak-light detection performance of perovskite PDs, including performance improvement and light absorption enhancement.

To systematically and quantitatively measure and compare the comprehensive performance of weak-light PDs, a series of performance metrics have been developed, such as responsivity (R), gain (G), specific detectivity (D*), and response time (τ). The comprehensive characterization of these devices in terms of these parameters provides a solid foundation for accurate performance evaluation and comparative analysis.

(a) Responsivity is an essential parameter that measures the ability of a PD in responding to a light signal. It is expressed as the current generated per unit optical power, and can be expressed as the following equation^[51]:

1R=Ip−IdP⋅A,

where I_p, I_d, and P represent the photocurrent, dark current, and light intensity, respectively, and A stands for the active area. The higher the responsivity, the more current the PD is able to generate when receiving the same optical power, indicating that the device is more sensitive in its response to the optical signal.

(b) Noise equivalent power (NEP) is a common metric to quantify the sensitivity of a PD or the power generated by a noise source. It represents the minimum detectable power per square root of bandwidth of a given PD, which is a measure of the weakest optical signal that can be detected^[52].

2NEP=in2¯1/2/R,

where i_n is the noise current of the PD and R is the responsivity. The internal noise level of the PD is one of the key factors affecting the NEP. As the device noise level decreases, the PD’s capability to detect weak-light signals improves, resulting in a lower NEP value. However, the higher the responsivity of the device, the stronger the signal the PD is able to produce when receiving the same optical power, thus reducing the NEP value to some extent.

(c) Specific detectivity (D*) is an indicator of the PD’s ability to detect weak signals. It integrates the sensitivity, noise level and spectral response of the PD and is an important basis for evaluating the performance of the PD. Its definition and calculation are based on the signal-to-noise ratio and noise equivalent power of the PD under specific conditions. It is given by^[53]:

3D*=A⋅fNEP,

where A and ƒ are active area and the electrical bandwidth, respectively. As mentioned above, NEP is a commonly used metric to quantify the sensitivity of a PD or the power generated by a noise source. Noise that limits the specific detection rate falls into three types: the shot noise from dark current, Johnson noise, and thermal fluctuation "flicker" noise. Under the assumption that shot noise is the dominant influencing factor, the expression for D* can be expressed as^[54]:

4D*=R⋅A2qId,

where I_d is the dark current, q is the absolute value of electron charge, and A represents the effective area.

(d) Gain is closely related to the responsivity of the device and typically refers to current gain, which measures the relationship between the output current and the input optical signal. It reflects the efficiency or sensitivity of the detector when converting an optical signal into an electrical signal. Higher gain indicates greater ability of the detector to detect weak-light signals and to produce measurable electrical signals under lower light conditions^[55].

5G=Rhceλ,

where R is responsivity, h is Planck’s constant, c represents the speed of light, e and λ stand for the elementary charge and the incident-light wavelength, respectively.

(e) Response time is a measure of how quickly a device responds to incident light. It represents the time delay between the reception of the optical signal by the PD and the generation of a stable electrical signal output. The shorter this time delay is, the faster the PD responds and the more sensitive it is to changes in the optical signal. The response time of a PD is usually obtained by experimental measurements, rather than being calculated directly by a simple formula. Generally, it is described by three common ways. ⅰ) The response time is defined as the time interval between 10% and 90% of the saturated photocurrent in the pulsed light irradiation period; ⅱ) the time required for the photocurrent to drop from a steady level to a photomodulation frequency of −3 dB; and ⅲ) the time constant extracted from a curve fitted to a biexponential function of the time response^[55].

PDs that can convert optical signals into electrical signals under weak-light conditions are crucial for fire warning, environmental monitoring^[56], etc. However, traditional PDs often struggle with low sensitivity and high noise under weak-light conditions, making it difficult to meet the demands for high precision and reliability. To further optimize the performance of perovskite weak-light detectors, both the performance improvement (including material optimization and device structure design) and light absorption enhancement are the current research focus. In terms of materials, precise modulation of the composition, crystal structure, and morphology of perovskite materials can effectively enhance their photoelectric conversion efficiency and stability. For device structure, optimizing the band structure and improving interface engineering are effective means to boost the performance of perovskite weak-light PDs^[57]. Furthermore, to achieve more efficient weak-light detection, designing and constructing special structures to enhance the capture and utilization of faint light signals has emerged as another crucial research direction. This innovative structural design enhances the detector’s sensitivity to weak light, allowing for a more efficient utilization of limited optical resources, thereby enabling high-precision detection even under weak-light conditions.

Perovskite materials, with their two main morphologies of polycrystalline thin films and single crystals, have shown a unique charm in the field of optoelectronics. In the delicate process of thin film preparation, the precise control of the crystallization process, especially the optimization of the annealing temperature and time, is the crucial factor in developing high-quality film morphology. This process directly determines the morphology quality of the film, which in turn profoundly affects the overall performance of the optoelectronic device. Therefore, optimizing the morphology of perovskite films to reduce defects such as pinholes and crystal boundaries can significantly reduce the recombination rate of the device and improve the photoelectric conversion efficiency^[58]. In contrast, perovskite single crystals exhibit significant advantages with their highly ordered structure, including low trap density, high mobility, and long diffusion length, which together make them ideal for applications pursuing high-speed and sensitive detection. With the deepening of research, the morphology of perovskite single crystals has evolved from three-dimensional (3D) large-size crystals to two-dimensional (2D) nanosheets, one-dimensional (1D) nanowires, and even zero-dimensional (0D) quantum dots. Not only this morphological diversity enriches the structural characteristics of perovskite materials, but also provides the possibility to achieve the modulation of different photoelectric response characteristics. For different morphologies of perovskite materials, we summarize the corresponding optimization methods and the corresponding photoresponsive properties.

Polycrystalline films: Regardless of whether the structure is horizontal or vertical type, the morphology of polycrystalline films is critical in contributing to the device performance. The presence of film defects can hinder the effective transport of photogenerated carriers and lead to the recombination of electrons and holes, thus weakening the detection ability of the PD or even leading to a short circuit of the device. Therefore, the morphology of perovskite films is one of the most critical factors determining the performance associated with PD devices. Improving the film quality and reducing defects are crucial for device performance. However, the preparation of uniformly dense perovskite films remains a challenge. Along this line, lots of efforts have been adopted by researchers to optimize the crystalline quality of the photoactive layer and thus reduce pinholes and defects^[59−61], such as improving spin-coating conditions and adding anti-solvent dropwise. Among them, spin-coating conditions including annealing temperature, annealing time, etc. have been widely used in most high-efficiency perovskite PDs. And it plays a crucial role in the growth of perovskite. The rotational speed of spin-coating determines the thickness of the film. The annealing temperature and annealing time, on the other hand, have an effect on the size and uniformity of the perovskite grains. Wang et al. explored the role of temperature on thin film crystallization and found that in the range of 60−90 °C, MAPbI₃ perovskite films have large domain sizes at higher temperatures^[47]. As the temperature increases, the surface becomes smoother and the photocurrent tends to be higher. However, when the temperature is too high, some cracks appear on the surface of the film and the current decreases. Within the same year, Wu et al. prepared the same MAPbI₃ perovskite films using a one-step solution method^[62]. During the preparation process, not only the annealing temperature (60−150 °C) was modified, but also the annealing time (10−20 min) was added as a variable to explore the most appropriate preparation conditions to achieve effective doping control. It was concluded in comparison that, on the one hand, the grain size of the films gradually expanded with the increase of annealing time and temperature. On the other hand, when the temperature is too high, more PbI₂ impurities appear. Therefore, the annealing conditions need to be precisely regulated in order to obtain the ideal spin-coating conditions, yielding dense, smooth, pinhole-free films of high quality. In addition to adjusting the operational part of spin-coating, the researchers employed an anti-solvent assisted crystallization strategy to obtain dense and smooth films. This strategy involves catalyzing heterogeneous nucleation near the end of spin-coating by rapidly adding a drop of an anti-solvent to create an instantaneous local supersaturation state. Yang’s group investigated the mechanical reconstruction process occurring in FAPbI₃ perovskite films during the post-treatment process^[63]. Through a combination of experiments and theoretical calculations, the treatment with isopropyl alcohol (IPA) on the surface of the perovskite films was able to significantly passivate the surface defects of the films, which stemmed from the fact that IPA reconstructs the FAPbI₃ surface and contributes to the anchoring of organic ammonium salts on the perovskite surface. Subsequently, surface post-treatment strategies were explored using distinct solvents to optimize inorganic Cs₂AgBiBr₆ perovskite films by Li et al.^[64]. The PDs were treated with five solvents respectively, IPA, diethyl sulfide (DES), chlorobenzene (CB), toluene (TOL), and anisole, and all of these solvents were found to have a positive effect on the device performance, with the best device performance after IPA solvent treatment. Comparing the devices before and after treatment, the R of the optimized device is doubled, and the calculated defect density of states is significantly reduced, the D* is up to 2.1 × 10¹² Jones, and the linear dynamic range (LDR) is 107 dB, indicating that the constructed self-powered PD effectively improves the sensitivity and thus facilitates the improvement of the weak-light detection capability. Anti-solvents that are commonly employed include chlorobenzene, toluene, methylene chloride^{[14, 31, 65]}, etc., which are known to be strongly toxic solvents. In order to respond to the national call for environmental protection and to further put green, safe and environmental protection into practice in experiments, an increasing number of researchers are looking for low-toxicity anti-solvents for substitution, such as methyl acetate (MA)^[66], especially in the preparation of highly homogeneous and dense lead-free copper-based perovskite films. The green MA has a high vapor pressure as well as solubility and is therefore widely used in the final step of spin-coating to obtain superior results. In the work of Zeng et al., the anti-solvent MA was added to the spin-coating of Cs₃Cu₂I₅ perovskite films, as shown in Figs. 4(a) and 4(b)^[67]. By comparing the top-view SEM images of Cs₃Cu₂I₅ perovskite films prepared with/without MA as the anti-solvent, it is clear that the dense and uniform films formed after the addition of the anti-solvent are of excellent quality compared to the perovskite films without the anti-solvent dropwise. Specifically, the Cs₃Cu₂I₅ perovskite film prepared without anti-solvent has numerous non-negligible pinholes with a grain size distribution in the range of 50 to 900 nm. As a comparison, the Cs₃Cu₂I₅ perovskite films prepared with anti-solvent completely covered the substrate and had a more concentrated range of grain sizes from 80 to 320 nm, and the photoluminescence quantum efficiency (PLQY) of the devices was increased by about 70%, which greatly improved the utilization of light absorption and enhanced the device performance. This illustrates that the dropwise addition of the anti-solvent during the spin-coating process can effectively accelerate the nucleation and crystal growth, thus improving the density and uniformity of the films.

These above results demonstrate that the morphology of perovskite crystals depends heavily on every single step of spin-coating. The optimal spin-coating conditions for different perovskite materials require variation and still need to be continuously explored in numerous experiments.

Single crystals: Polycrystalline perovskite films tend to be relatively unstable in air, and water molecules as well as oxygen can easily enter the interior of the film through grain boundaries or defects, triggering the degradation of the perovskite material, which leads to poor long-term stability of the perovskite film. Compared with rapidly degrading, fragile and boundary-rich films, single crystals show excellent stability in humidity, light and high temperature. Thus, we started from exploring 3D large-size perovskite single crystals and further explored 2D nanosheets, 1D nanowires, and 0D quantum dots. As shown in Fig. 4(c), Zhang et al. obtained MAPbBr₃ single crystals with regular shapes using a simple room temperature crystallization method^[68]. A smooth-surface, centimetre-sized single crystal with large dimensions was obtained by the gradual growth of single crystals in the precursor solution from Figs. 4(c) (ⅰ) and (ⅱ) over a week. It is this slow growth process that allows the crystals to have more time for orderly arrangement and crystallization, thus improving the quality of the single crystals. In addition, inverse temperature crystallization and anti-solvent vapor-assisted crystallization are also common means of preparing single crystals^[73]. Inverse temperature crystallization can usually form medium-sized perovskite single crystals in a short time. While antisolvent vapor assisted crystallization can form a large number of nuclei in a short time and grow them gradually.

Despite the renown of 3D perovskite single crystals for their smooth surfaces and exceptional electron mobility, their inherent rigidity significantly hinders effective integration with flexible substrates. In response, more and more researchers have shifted their focus to 2D perovskite crystals, which not only exhibit remarkable enhancements in optical response and photocarrier lifetime but also demonstrate high compatibility with flexible substrates due to their ultrathin form. In 2016, Song et al. first reported the synthesis of 2D CsPbBr₃ nanosheets with a thickness of only 3.3 nm via a solution-phase synthesis method^[74]. Their 2D nature renders them suitable for the fabrication of flexible PDs, which, after undergoing over 10 000 cycles of bending tests, still maintain outstanding stability, fully demonstrating their potential for practical applications. In 2021, Huang et al. first synthesized lead-free 2D halide Cs₂AgBiBr₆ nanosheets through a colloidal synthesis method^[69]. As shown in Fig. 4(d), these nanosheets possess a thickness ranging from 3−5 nm and a lateral size of approximately 200 nm. Remarkably, by adjusting the precursor concentration, they achieved a morphological transformation of Cs₂AgBiBr₆ from 0D nanocubes to 2D nanosheets, providing novel insights and methodologies for material dimensionality control.

The solution method, a commonly used technique for the preparation of 1D perovskite microwires, is centred on the precise regulation of the crystalline growth process of perovskite materials along a specific axial direction in order to shape the microwire morphology. Within this process, as the perovskite precursor solution fills the voids within the template, capillary forces act as a driving force, gradually propelling the solution to flow along the preset template with a specific morphology. Furthermore, perovskites with anisotropic crystal structures are naturally inclined to grow along a particular crystal phase, thus displaying an obvious one-dimensional feature on the macroscopic scale. By regulating the temperature of the heating plate, the solvent within the solution gradually evaporates, enabling the material to grow orderly along the predetermined direction, finally obtaining the micron lines with the morphology and size meeting the preset criteria. Li et al. employed a template-assisted strategy to synthesize MAPbBr₃ microwires^[70]. The single-crystal microwire arrays obtained through this process exhibit high crystal quality and uniform morphology, as shown in Fig. 4(e). The single crystal microwire arrays (SCMWAs) are precisely aligned with the templates, allowing for the alteration of the final MAPbBr₃ SCMWAs morphology by simply modifying the templates. By employing templates with periodically curved structures, curved MAPbBr₃ SCMWAs can be achieved. Both the linear and curved arrays exhibit nearly perfect morphologies, with uniform heights and widths. Notably, for the curved microwires, there are no observable cracks or other defects even at the inflection points. Through an innovative in-situ encapsulation technique, they constructed a protective hydrophobic molecular layer on the surface of the microwires, significantly enhancing the stability of MAPbBr₃ SCMWAs in ambient air. Even after prolonged exposure to air, their performance remains unaltered for over a year. Furthermore, their group extended their technique to the fabrication of curved perovskite microwire crystal arrays^[75], achieving a morphological transformation from planar to complex curved surfaces and successfully integrating them into flexible imaging systems, demonstrating immense potential for practical applications. Experimental results showcase that after storage in ambient air for 297 days and even longer periods up to 831 days, the device performance retains 97% and 85% of its initial values, respectively, underlining its remarkable long-term stability. Moreover, the group delved into the construction of heterojunctions on individual microwires, utilizing a two-step imprinting method to successfully fabricate MAPbI₃−MAPbBr₃ microwire lateral heterojunctions^[71], as shown in Figs. 4(f) and 4(g). In this heterojunction, the green and red colors at the two ends represent MAPbBr₃ and MAPbI₃, respectively, and the interface is where the two crystals intermingle, forming an in-plane lateral heterojunction structure. These heterojunctions exhibit exceptional optoelectronic properties, boasting responsivity and detectivity values as high as 1207 A·W⁻¹ and 2.78 × 10¹³ Jones, respectively. Notably, the resulting PD not only possesses outstanding flexibility, retaining 83.3% of its original performance after 3000 bending cycles, but also maintains over 90% of its initial photoresponse even after 100 days of exposure to air.

When the size of perovskite nanocrystals (NCs) is reduced below the Bohr radius, their extreme small size imparts these nanocrystals with significant quantum confinement effects^[76]. This effect greatly enhances the size and shape dependence of quantum dots in optical and electrical properties, providing unprecedented flexibility in regulating these properties^[77]. Ma et al. successfully synthesized Cs₃Bi₂X₉ quantum dots using a facile ligand-assisted recrystallization method^[72], which does not require high temperatures, high costs, inert gas protection, or complex injection conditions. The Cs₃Bi₂Br₉ quantum dots prepared by this method were ingeniously encapsulated in a BiOBr matrix, forming Cs₃Bi₂Br₉/BiOBr nanocomposites. This composite material not only retains the unique quantum effects of quantum dots, but the presence of the matrix also enables the material to exhibit superior stability and performance. As shown in the high-resolution TEM image in Fig. 4(h), the distribution of Cs₃Bi₂Br₉ quantum dots within the BiOBr matrix and the (202) crystal plane of Cs₃Bi₂Br₉ can be easily observed, with the interplanar distance of this crystal plane precisely measured as 2.82 Å.

To enhance the performance of PDs, a rational device structure is paramount. By optimizing energy level arrangements and implementing interface engineering strategies, the performance of perovskite PDs for weak-light can be significantly improved. Inserting transport layers or blocking layers between the perovskite material and electrodes not only optimizes the band alignment structure of the device but also enhances charge collection efficiency and device stability. Specifically, the introduction of p-type and n-type transport layers can reduce interface reactions and facilitate effective carrier separation, while simultaneously blocking reverse charge injection, inhibiting ion migration, and lowering dark current.

Zhou et al. constructed a CuI/CsCu₂I₃/GaN device with staggered band alignment (Fig. 5(a))^[78], which forms a typical type-Ⅱ heterojunction at the material interface, accompanied by a built-in electric field pointing from GaN to CuI (Fig. 5(b)). Under illumination, this built-in electric field accelerates the separation of photogenerated carriers, thereby reducing their recombination and enabling the heterojunction device to exhibit an extremely low dark current level (3.6 × 10⁻¹⁰ A) at zero bias. As early as 2014, Dou et al. constructed hole transporting layer (HTL)/perovskite/electron transporting layer (ETL) devices (Figs. 5(c) and 5(d)) to investigate the suppression effect of hole blocking layers on dark current^[18]. They designed three detectors with different configurations (PD1 without a blocking layer, PD2 with a BCP blocking layer, PD3 with a PFN blocking layer) and found through J−V curve analysis (Fig. 5(e)) that the introduction of hole blocking layers significantly reduced dark current and improved the rectification ratio. Notably, PFN further optimized electron injection and extraction processes by lowering the metal cathode work function via a surface dipole mechanism, exhibiting superior performance. However, constructing ideal transport layers with matched energy levels, smooth surfaces, and dense structures still poses technical challenges. To address this, Cao et al. proposed an innovative strategy of utilizing gradient O-doped CdS nanorod arrays in conjunction with perovskite to construct PDs^[79]. They introduced gradient built-in bands at the charge extraction layer and perovskite interface, which not only strengthened carrier separation but also effectively suppressed electron backflow and carrier recombination, ultimately achieving a specific detectivity as high as 2.1 × 10¹³ Jones. Fig. 5(f) illustrates the fabrication process of this gradient PD, where CdS nanorod arrays are grown on FTO glass via hydrothermal synthesis, followed by heat treatment in air. During this process, oxygen in the air gradually diffuses from the surface to the interior of the CdS nanorods, forming gradient O-doped CdS nanorods. In gradient O-doped CdS, as oxygen concentration increases radially, all energy levels including E_C, E_f, and E_V rise, causing continuous bending of the embedded energy bands, resulting in the band alignment shown in Fig. 5(g).

Moreover, interface modification is another crucial approach to enhancing device performance. Qiu et al.’s research revealed that although different alkylammonium salt interlayer (ASI) modifications did not alter the lattice structure of perovskite films or introduce new phases (Fig. 5(h))^[80], they significantly promoted grain growth and improved film quality. Simultaneously, selecting appropriate electrode materials can optimize the internal electric field distribution within the device, enhancing carrier collection efficiency. Chen et al. utilized surface-modified MXenes as electrodes to replace traditional gold electrodes^[81], significantly improving charge separation efficiency by adjusting the Schottky barrier (Fig. 5(i) shows the energy band diagram at thermal equilibrium), thereby enhancing the rectification performance of the device. When the optimal MXene-PEIE was used as the electrode, the rectification ratio of the device reached as high as 16136 (@±2 V).

In addition to enhancing the performance of perovskite PDs through optimized preparation processes and device structures to achieve weak-light detection, employing special designs to improve the utilization efficiency of weak-light is also an extremely effective strategy. Inspired by the remarkable night vision capability of owls in nature, Guo et al. have designed a biomimetic perovskite PD featuring a dense array of micro-hemispherical structures on its surface^[82], exhibiting extraordinary sensitivity to weak light. Compared to devices without this structure, the R and D* of this PD soared by 16.4 and 11.7 times, respectively. The SEM image in Figs. 6(a) and 6(b) clearly demonstrates the high-quality crystalline morphology of these micro-hemispherical structures, which act like miniature light-condensing lenses, significantly facilitating light capture and convergence, thereby enhancing the PD’s sensitivity to faint light signals. Furthermore, Li et al. reported a MAPbBr₃ crystal with a moiré lattice structure, which was prepared by utilizing two soft templates of nanograting structures with a relative rotation angle^[83]. This unique structure (Fig. 6(c)) can enhance the light absorption of perovskite crystals, with the PD exhibiting a responsivity of 1026.5 A·W⁻¹.

Moreover, the ingenious application of plasmonic technology has paved a new way for enhancing the photoelectric performance of perovskites. By fine-tuning surface composition, suppressing charge recombination, and introducing plasmonic nanostructures such as metal nanoparticles and nanorods, it not only optimizes interfacial charge collection but also significantly enhances light absorption and scattering through the localized surface plasmon resonance (LSPR) effect, thereby dramatically boosting photoelectric performance^[86]. Li et al. ingeniously hybridized metal nanoparticles with anodic aluminum oxide (AAO) to form plasmonic nanostructures, realizing efficient light utilization through a spatially extended light confinement strategy^[87]. Meanwhile, the bowtie nanoantenna (BNA) array structure (Fig. 6(d)) constructed by Wang’s team leveraged out-of-plane near-field coupling effects to elevate the PD’s responsivity by 2962 times and catapult the D* value two orders of magnitude higher^[84]. Its LSPR coupling mechanism substantially intensified the electric field, broadened the light detection bandwidth, and enabled efficient detection from visible to near-infrared light. The finite-difference time-domain (FDTD) simulation results (Figs. 6(e) and 6(f)) visually showcase the electric field distribution within the BNA structure, particularly the pronounced enhancement at the bowtie gap and the tips and ridges of the metal−insulator−metal (MIM) structure, further validating the pivotal role of plasmonic nanostructures in enhancing PD’s performance. Compared to layered nanostructures solely relying on simple nano-grating layouts, the layered design incorporating cross-nanogratings (CGs) exhibits notably enhanced light-trapping capabilities. Lee et al. innovatively introduced a layered plasmonic structure integrating CGs and nanoposts (NPs)^[85], fabricated through the advanced technologies of block copolymer lithography and nanoimprint lithography, as depicted in Fig. 6(g). The perovskite PD based on this structure exhibits a responsivity of up to 580 mA·W⁻¹ and a detectivity of 3.2 × 10¹² Jones. Compared with similar detectors without integrated plasmonic nanostructures, these two indicators have increased by 420% and 990%, respectively, demonstrating a significant leap in photoelectric conversion efficiency. To delve into the underlying physical mechanisms behind this performance enhancement, researchers employed FDTD theoretical calculations (Fig. 6(h)) to uncover the efficient light-trapping effect achieved by the layered plasmonic mode within the perovskite layer.

With the vigorous development of emerging technologies such as autonomous driving and biosensors, the demand for flexible and bendable PDs is growing rapidly. Traditional PDs primarily employ rigid silicon materials as sensing units. While they exhibit outstanding performance, their inherent brittleness and inflexibility limit their application scope in flexible optoelectronic devices. Consequently, there is an urgent need for a superior alternative to achieve better integration with human skin while ensuring data fidelity and accuracy. On the other hand, perovskite materials, as an emerging class of optoelectronic materials, have opened up new possibilities for wearable health monitoring technologies due to their exceptional optoelectronic properties, low-temperature solution processability, and remarkable mechanical flexibility. With advancements in the preparation techniques of perovskite materials and their excellent compatibility with arbitrary substrates, perovskite optoelectronic devices are expected to play a key role in wearable devices and intelligent monitoring sensor applications.

Wearable health monitoring devices, with their increasingly refined miniaturization and convenience, are quietly integrating into people’s daily lives. By leveraging emerging perovskite materials as efficient light-absorbing layers in optoelectronic devices, these devices have not only achieved real-time monitoring of traditional physiological indicators such as heart rate and blood pressure but have also explored their potential applications in skin health monitoring, providing robust technical support for personalized health management. Meanwhile, the rapid development of Bluetooth technology has enabled seamless connectivity between wearable devices and intelligent terminals like smartphones. In light of this, this section will focus on two core areas: the advancements in photoplethysmography (PPG) signal monitoring technology and the innovative applications of integrated smart tools in wearable devices, comprehensively reviewing the latest progress and future trends of wearable health monitoring technology.

The flexible organic PPG sensor, an innovative physiological monitoring technology, achieves real-time monitoring of critical physiological parameters such as heart rate and blood oxygen saturation by precisely capturing subtle changes in blood volume beneath the skin. A light-emitting diode (LED) is used as a light source to emit light of a specific wavelength, which penetrates the surface of the skin and is selectively absorbed by the hemoglobin in the deeper blood vessels. As the heart contracts and relaxes periodically, the blood volume in the arteries changes, leading to variations in the attenuation of light in the blood. Subsequently, highly sensitive PDs analyze these reflected or transmitted light signals to calculate key physiological data like blood oxygen saturation. Traditional commercial pulse oximeter sensors, often designed as fingertip clips with commercial LEDs as stable light sources and rigid silicon materials as sensing elements, are widely used in clinical and home health monitoring but have room for improvement in flexibility and comfort. Based on the relative positions of the light source and detector, pulse oximeter sensors can be categorized into reflective and transmissive types^[88]. In the transmissive mode (Fig. 7(a)), the light source and detector are placed on opposite sides of the measurement area, requiring light to fully penetrate the measured site (such as the earlobe or finger with good light transmission) for effective reception. This approach imposes certain requirements on the light transmission properties of the measurement site. In contrast, the reflective mode is more flexible (Fig. 7(b)), with both the light source and detector located on the same side of the measurement area, capturing and analyzing backscattered light signals from the object. This design simplifies the sensor structure and significantly expands its application scenarios, enabling flexible organic optoelectronic PPG sensors to function in more diverse environments and conditions, providing users with a more convenient and comfortable health monitoring experience.

Transmission mode significantly reduces the reflection and scattering effects of light on the skin surface and its surroundings, which effectively reduces the interference of external light fluctuations and skin surface characteristics on the signal detection. By placing the light source and detector on opposite sides of the object, the technique allows the detector to focus on capturing the transmitted light, which in turn places higher demands on the light sensitivity of the detector. Polycrystalline metal halide perovskite materials have long plagued the accurate recognition of PPG signals due to high dark currents caused by defects and ion migration. In 2023, Tang et al. proposed an electric field modulation strategy^[25], successfully developing a high signal-to-noise ratio wearable PPG sensor (Figs. 7(c) and 7(d)) that effectively suppresses baseline drift, enabling the acquisition of high-fidelity PPG signals even in weak-light environments. When the incident light intensity decreases to 2 mW·cm⁻², the blood pulse signal is obscured by baseline drift and cannot be identified (Fig. 7(g) (ⅰ)). In contrast, under 0.1 V control voltage modulation, high-fidelity PPG signals can be obtained regardless of light intensity (Figs. 7(e)−7(g) (ⅱ)). Subsequently, in 2024, Liu et al. prepared FASnI₃-CNI perovskite films with lower non-radiative recombination and defect density through additive engineering, enabling zero-power operation of self-powered PDs and successful application in real-time monitoring of human heart rate^[94], further advancing the use of flexible non-toxic materials in health monitoring. Moreover, Liu et al. innovatively applied mixed Sn−Pb perovskite materials to a wearable remote health monitoring system (RWHM) in the transmissive mode^[89], integrating health monitoring and remote optical communication modules to achieve long-distance, high-precision monitoring of physiological parameters such as heart rate. In the health monitoring module, the PD serves as a high-precision biosensor for PPG testing, capturing pulse fluctuations non-invasively and accurately calculating heart rate. Meanwhile, the remote optical communication module intelligently modulates the pulse waveform onto blue LED light signals, serving as the transmission source for optical communication, enabling wireless data transmission and analysis (Fig. 7(h)), demonstrating stable monitoring capabilities in complex environments. Whether in resting states or during post-exercise recovery, the pulse signals captured by this system (Figs. 7(i) and 7(j)) exhibit high sensitivity and reliable detection rates, fully demonstrating its potential for stable and accurate health monitoring in complex environments.

However, the thickness of the measurement site significantly impacts the transmissive mode, and it is sensitive to changes in ambient light. Strong or unstable ambient light may interfere with the transmitted light signals. In contrast, the reflective mode is applicable to most body parts, offering stronger versatility. Wu et al. prepared a fingertip patch-type optoelectronic system (Fig. 7(k)) by integrating flexible perovskite PDs with all-inorganic LEDs, enabling real-time monitoring of PPG signals and assessment of pulse rate and finger swelling, providing a new avenue for early disease prevention and diagnosis (Figs. 7(l)−7(o))^[90]. Xu’s team, on the other hand, utilized all-inorganic perovskite films prepared through vapor deposition technology to create highly stable pulse monitoring sensors (Fig. 7(p))^[91]. As shown in Fig. 7(q), these sensors maintain signal stability even under extremely weak light conditions (0.055 mW·cm⁻²), injecting new vitality into the development of wearable optoelectronic devices.

In conclusion, flexible optoelectronic PPG sensors, whether operating in transmission or reflection mode, have demonstrated immense application potential and promising development prospects, gradually emerging as a significant force in the future of health monitoring. Furthermore, as an efficient light-absorbing layer, the perovskite material is not only limited to the accurate monitoring of PPG signals, but also shows multiple applications in daily health monitoring, including real-time monitoring of UV intensity, simulation of human skin’s sunburn response and post-sunburn restoration, etc., which provides a full range of technical support for personalized health management.

Zhou et al. prepared a flexible PD based on MAPbCl₃ and combined it with a flexible printed circuit board (FPCB) to obtain a wearable wristband (Fig. 7(s)) for real-time UV monitoring^[92]. As depicted in Fig. 7(r), this flexible sensor system can display the current UV radiation intensity in real-time and record it on a smartphone. However, this PD cannot simulate the cumulative damage of UV rays to the skin and can only give the current real-time UV intensity. Based on this, Shen’s team prepared flexible wearable optoelectronic synapses (Fig. 7(u)) based on PEA₂SnI₄ perovskite recently^[93]. The schematic diagram is shown in Fig. 7(t), when exposed to light, the device’s photocurrent cannot immediately reach saturation. Instead, its current gradually increases with the increase in illumination time. This characteristic makes it possible to use continuous illumination to simulate the cumulative damage of UV rays to the skin.

Due to their unique optoelectronic properties and mechanical flexibility, perovskite materials are widely used in wearable health monitoring devices. By integrating these materials into sensors, real-time monitoring of a variety of physiological parameters such as heart rate, blood pressure, oxygen saturation, as well as the degree of skin damage caused by external UV rays can be achieved. This real-time monitoring capability enables users to stay informed about their health status and take appropriate measures when necessary. Therefore, wearable health monitoring based on perovskites provides solid technical support for people to enjoy a healthier and more convenient lifestyle.

In 2017, Fang et al. innovatively demonstrated a simple and intriguing PD that leveraged graphite electrodes traced by a pencil on paper and MAPbI₃ droplets as the light-absorbing layer^[35]. A 3D stereoscopic arrayed perovskite device was constructed by folding along predefined creases, as depicted in Fig. 8(a). By establishing a Cartesian coordinate system (Figs. 8(b) and 8(c)), this 3D PD array could respond to incident light from any direction, enabling the determination of light incidence angles through signal processing, thereby facilitating its application in spatial recognition. Additionally, Leung et al. constructed a flexible and transparent self-powered PD based on MAPbI₃ material, capable of functioning under omnidirectional illumination^[95]. Notably, the device exhibited varying performance under different environmental conditions, including sunny, cloudy, and indoor lighting (Fig. 8(d)), indicating its ability to discern varying light intensities, offering new directions for next-generation optical communication, sensing, and imaging applications. As urbanization accelerates, high-rise buildings such as residential complexes, research laboratories, and large malls have become densely populated areas with frequent activities. The proliferation of electrical facilities and flammable materials significantly elevates the risk and potential damage of fires. Consequently, deploying efficient automatic fire alarm systems on each floor of these buildings^{[97, 98]} to achieve early fire detection and rapid response is crucial for reducing losses and ensuring safety. Among various fire alarm devices, flame detectors stand out due to their unique advantages as light-sensitive fire detectors that operate based on the infrared or ultraviolet characteristics emitted during flame combustion. Zhang et al. developed a flame monitoring system utilizing a UV PD, which effectively eliminated visible light interference through a narrowband pass filter^[3], significantly reducing false alarms. However, the rigidity of its substrate limited its compatibility with flexible structures, thereby restricting its application scenarios. Ingeniously, Lu et al. employed origami techniques to design a flexible PD array comprising 40 pixel units that could seamlessly adhere to hemispherical structures^[96]. The array, with eight branches meticulously labeled A to H in a counterclockwise direction, each housing five evenly spaced PDs, formed a monitoring network spanning a wide-angle range from 0° to 180° (Fig. 8(e) (ⅰ)), enabling precise capture and identification of multiple flame signals. When a flame radiation intensity reached 220 µW∙cm⁻² and approached the A5 position in the array, the system promptly responded, with the current distribution map visually showcasing a significant enhancement in photocurrent around A5 (Fig. 8(e) (ⅱ)), while the rest of the region maintained a stable dark current state. This distinct current contrast provided a solid basis for researchers to map the 3D current distribution (Fig. 8(f)), pinpointing the exact flame location. Equally impressive, the system excelled in multi-flame detection, as shown in Fig. 8(g). During experiments, flames from three different directions were simultaneously positioned around the array, and by analyzing the photocurrent data, the device accurately identified flames at the key locations of A5, C2, and F4, demonstrating its outstanding detection capabilities and broad applicability.

The low cost, ease of processing, and scalability of flexible perovskite PDs offer robust guarantees for their widespread adoption in environmental monitoring. With continuous technological advancements and further cost reductions, this novel PD is poised to become a mainstream technology in the field of environmental monitoring, contributing to the construction of a greener and more sustainable ecological environment.

In the profound exploration of optical imaging technology, PDs, as the core technical components, can be distinctly categorized into two major types based on their imaging mechanisms: transmission imaging and reflection imaging. The transmission imaging mechanism heavily relies on the PD’s exquisite ability to perceive and quantitatively convert transmitted light rays. The high sensitivity and low noise characteristics exhibited by PDs have significantly contributed to the widespread adoption and deepened development of transmission imaging technology in cutting-edge fields such as biomedical research and materials science exploration. Meanwhile, reflection imaging also depends on the PD’s efficient response and precise conversion of reflected light signals. The geometric shape, surface texture, and color information inherent in reflected light are transformed into electrical signals by the PD, which are then subjected to intricate processing and image reconstruction by digital systems, ultimately resulting in an intuitive and informative visual representation. It is noteworthy that early imaging research was predominantly focused on single-pixel imaging technology, which, due to its unique high spectral resolution and exceptional low background noise performance, demonstrated irreplaceable advantages in areas such as detection under weak-light conditions and high-precision spectral analysis. Nevertheless, the inherent limitation of single-pixel imaging lies in its relatively slow imaging speed. In response, array imaging technology emerged^{[99, 100]}, which integrates high-density, high-performance PD arrays to achieve parallel and high-speed acquisition of optical signals and instant reconstruction of image data. Consequently, this technology has significantly enhanced imaging efficiency and resolution, becoming a pillar technology in fields like high-definition video capture and dynamic scene monitoring.

Li and his team ingeniously constructed a transmission imaging system based on the SnO₂/Cs₂AgBiBr₆ heterojunction PD^[64]. The design principle of this system is illustrated in Fig. 9(a), which integrates a laser, a precise 2D scanning platform, an object to be imaged, a source meter, and the core PD. Within the system, a Chinese character pattern approximately 1 cm in size serves as the imaging template, and its spatial resolution (x, y) fineness is directly determined by the precise manipulation of the imaging sample by the 2D scanning platform. During the imaging process, the transparent regions allow light to penetrate and trigger the PD to record photocurrent signals, while the non-transparent regions block the light, resulting in the recording of dark current signals. Figs. 9(b) and 9(c) respectively demonstrate the imaging effects of this system under extremely weak light intensities (0.5 and 5 μW·cm⁻²), significantly validating the exceptional imaging capability of this heterojunction PD in weak-light environments. Dong et al. took a different approach by adopting 2D perovskite (PA)₂PbBr₄ single crystals as the active material for planar PDs, developing a highly sensitive UV detector with an ultra-low detection threshold^[103]. The clarity and resolution of its imaging are attributed to the remarkable performance of the PD, including an extremely weak dark current (as weak as 0.735 pA) and significant weak light response capabilities even under extremely weak illumination (5.49 nW·cm⁻²).

The architecture foundation of a reflection imaging system is similar to that of a transmission imaging system, with the primary difference lying in the fact that the former captures light reflected from the surface of an object, while the latter directly receives incident light. As depicted in Fig. 9(d), a reflection imaging system integrates critical components such as lasers, precise two-dimensional scanning platforms, imaging samples, source meters, and PDs. During the imaging process, the laser beam is focused by an optical lens and shone onto the surface of the imaging sample. Subsequently, diffuse reflection occurs on the sample surface, with the intensity characteristics of the reflected light correlated to the sample’s absorption coefficient and surface morphology. Wu et al. conducted a comparative study exploring the potential of flexible MAPbI₃ nanowire PDs in the field of diffuse reflection imaging, contrasting them with commercial silicon photodiodes (S2386)^[101]. Their results revealed that the flexible detectors exhibited superior performance in diffuse reflection imaging. This discovery not only opens up a new path for constructing high-performance, stable, and flexible perovskite PDs but also provides strong support for accelerating their commercialization. Figs. 9(e) and 9(f) respectively show the "butterfly outline" diffuse reflection images detected by the commercial silicon photodiode S2386 and the constructed flexible PD. Compared to the commercial silicon photodiode S2386, this flexible PD displays a clearer and more distinct diffuse reflection image, indicating its higher capability in diffuse reflection imaging compared to the commercial counterpart. Meanwhile, Zhang et al. conducted an in-depth study on the application of MAPbBr₃ MWAs in single-pixel reflection imaging technology, thoroughly comparing them with commercial silicon photodiode single-pixel detectors (Fig. 9(g))^[102]. As demonstrated in the reconstruction results shown in Fig. 9(h), both detectors effectively capture the reflected light signals from color objects. However, when distinguishing complex patterns like the "2" signal, the MAPbBr₃ MWAs single-pixel detector exhibits superior accuracy, particularly in the blue-biased side where its performance is particularly outstanding. Moreover, this detector achieves an ultra-high wavelength resolution of approximately 20 nm in reconstructing reflected color object images, far surpassing the color information capture capability of commercial silicon photodiode PDs. Critically, its minuscule effective sensing area of 1.3 × 10⁻³ mm² represents a significant downsizing compared to the 75.4 mm² of commercial silicon photodiodes, portending vast application prospects in compact and efficient imaging domains.

Single-pixel imaging systems are often limited by their imaging method of sequential measurement or scanning, resulting in slower imaging rates that struggle to meet the demands of real-time dynamic imaging. In contrast, array imaging systems contain multiple pixels, each capable of simultaneously receiving and processing optical signals, leading to significantly faster imaging rates and enabling real-time dynamic imaging. Luo’s team used MAPbBr₃ PDs to construct a 6 × 6 array device, which was able to correctly record the motion of the light spot, and was able to recognize individual characters^[104]. However, the response speed of traditional PDs and electrical crosstalk between arrays have been critical factors limiting their performance. To address this, Zhan et al. innovatively designed a 4 × 4 array with a perovskite photodiode-blocking diode (PIN-BD) crossbar structure^[105]. This design ingeniously reduces electrical crosstalk within the array to just 8.0%, achieving high-precision capture of static images and smooth capture of real-time dynamic images.

As thoroughly examined in this article, the diverse preparation methods of perovskite materials and their exceptional compatibility with flexible, stretchable substrates have fueled the rapid advancement of flexible weak-light detector technology, revealing unprecedented potential in various fields. With the relentless innovation of artificial intelligence technology, these high-performance optoelectronic devices can seamlessly integrate into flexible wearable devices. However, to fully realize this revolutionary transformation, it is imperative to foster deep integration and collaborative innovation across multiple disciplines, including materials science, flexible electronics technology, and artificial intelligence algorithms. In this process, the challenges cannot be overlooked: transitioning from small-scale laboratory devices to large-scale industrial production, enhancing the long-term stability of perovskite materials, and ensuring the repeatability of device fabrication processes and the uniformity of products are all crucial technical hurdles that must be overcome urgently.

Commercial silicon-based solar cells generally enjoy a lifespan of approximately 20 years^[106], while commercial optoelectronic devices such as smoke detectors are typically capable of stable operation for around 10 years. Semiconductor materials inevitably undergo an aging process when exposed to air over time, a process that directly impairs device performance and ultimately affects their normal functionality. In contrast, current perovskite PDs exhibit limited stability in air, constrained to short-term durations measured in months, significantly lagging behind commercial standards. This suggests that high-performance PDs fabricated from perovskite materials must surmount numerous challenges before achieving widespread adoption. Indeed, when assessing PD stability, it is imperative to thoroughly consider their performance under various extreme operating environments, including but not limited to humid and high-temperature conditions, as these factors directly correlate with the device’s sustained operational capability and the stability of its optoelectronic performance parameters. Against this backdrop, perovskite PDs must demonstrate remarkable adaptability to maintain stable and superior performance under all types of extreme conditions.

Solutions to the stability are deeply researched and investigated from two main aspects: extrinsic and intrinsic factors. Extrinsic factors are primarily addressed through advanced encapsulation techniques, such as using parylene-C^[107] and polydimethylsiloxane^{[3, 83]}, to encapsulate devices, mitigating the ingress of water and oxygen and significantly enhancing their stability. Nevertheless, intrinsic stability issues within devices cannot be fully resolved solely through external encapsulation. Therefore, enhancing the intrinsic stability of materials through material optimization is a powerful way to further improve the stability of devices. Research conducted by the Kandasamy Prabakar team in 2018 revealed that perovskite material undergoes self-degradation processes even in dark and vacuum conditions^[108]. Key factors influencing traditional MAPbI₃ degradation include perovskite structure and ionic defects. Thus, solutions should focus on the ABX₃ structure of perovskites to ensure structural stability while continually enhancing optoelectronic performance. To date, the most common A-site equivalent organic cation substitute is FA⁺, which is slightly larger than the methylammonium ion (MA⁺) and forms stronger hydrogen bonds with I⁻, thereby imparting better stability, particularly under high-temperature conditions^[109]. Additionally, inorganic perovskites have garnered significant research attention in recent years^{[110, 111]}, with inorganic cations (e.g., Cs⁺) replacing organic cations (MA⁺) to markedly alter grain size and thermal stability^[112]. However, a more formidable challenge lies in enhancing PD stability against moisture and light. Studies have shown that 2D/3D hybrid perovskite systems exhibit remarkable humidity stability^[113]. By strategically inserting large organic cations into the perovskite structure as spacers between inorganic frameworks, not only is the hydrophobicity of the hybrid material significantly enhanced, but also a close link between the humidity stability and structural/thermodynamic stability of hybrid perovskites is revealed^{[114, 115]}. To further optimize this property, an in-depth exploration of the complex chemical interaction mechanisms between large cations and inorganic frameworks is required, providing a solid theoretical foundation and technical support for enhancing the moisture resistance of these materials.

In addition, due to the toxicity concerns associated with lead perovskite materials, researchers have been striving in recent years to ensure the environmental friendliness and safety of these materials while preserving their exceptional optoelectronic properties. They have delved deeply into the electronic configurations of typical lead halide perovskites and, based on this understanding, explored strategies to replace lead with homovalent and heterovalent elements. This not only retains the inherent advantages of perovskites but also enhances their stability and optoelectronic performance to a certain extent. Alternatively, a single heterovalent cation can replace Pb²⁺ while introducing mixed-valence anions to balance the charge, offering the flexibility to adjust the electronic structure and optical properties of perovskites more precisely. The vacancy-formation strategy represents another effective heterovalent substitution method, involving the creation of vacancies within the perovskite structure to maintain charge neutrality. The highly ordered arrangement of these vacancies in the lattice contributes to further improvements in the optoelectronic properties and stability of perovskites.

Selecting appropriate lead-free perovskite materials and implementing reasonable encapsulation strategies can effectively elevate the performance and stability of these lead-free perovskite PDs, thereby advancing their widespread adoption in optoelectronic technologies.

Despite the remarkable potential demonstrated by perovskite materials in the field of photoelectric conversion, particularly in weak-light detectors that have been specifically optimized to rival the performance of mature detectors on the market^{[101, 102]}, significant strides are still necessary to bridge the gap from laboratory research to productization and commercialization, and ultimately achieve the ambitious goal of replacing commercial silicon-based detectors. To accelerate the commercialization of perovskite PDs, the primary task lies in broadening and deepening their production scale while enhancing the reproducibility and consistency of manufacturing processes. In this endeavor, actively exploring and adopting low-cost, high-efficiency manufacturing techniques, such as spray coating and spin coating, is pivotal. These techniques are anticipated to substantially reduce production costs and facilitate rapid, large-scale production of perovskite PDs. Additionally, specific process optimization tailored to different perovskite morphologies is essential to ensure that every production batch maintains exceptional and stable performance. Furthermore, in-depth optimization at the device structural design level is necessary, involving precise modulation of material arrangements, interface properties, and functional layer configurations to significantly enhance the comprehensive performance of PDs, including sensitivity, response time, specific detectivity, and long-term stability, thereby laying a solid foundation for their commercial applications.

With the continuous development of artificial intelligence technology, perovskite PDs are gradually venturing into the innovative field of flexible and stretchable devices, paving a practical path for remote diagnosis and treatment in precision medicine. These flexible devices, deeply integrated with robots and artificial electronic skin equipment, exhibit great potential. Their synergistic effects not only significantly elevate the intelligence of machinery but also foster seamless integration of technology into daily lives, bringing science and technology closer to us.

In conclusion, perovskite materials hold a bright future in the optoelectronic field. Through in-depth optimization of materials and structures, the productization and commercialization of perovskite PDs will accelerate, contributing significantly to technological progress and social development. Looking ahead, with the continuous maturation and expanded applications of artificial intelligence technology, we believe that these flexible PDs will show their unique charm and value in even more fields, emerging as a vital force driving social progress and development.