The innovation of functional materials with tunable optoelectronic properties will take essential positions in the development of fundamental and applied research fields. Metal halide perovskite materials with a typical crystal structure such as CaTiO3 would evolve into an outstanding semiconductor counterpart to surpass all traditional materials in the optoelectronic field [1,2]. The general chemical formula of perovskites is ABX3, where “A” and “B” are two cations of very different sizes and “X” is an anion that bonds to both “A” and “B.” Progress has been achieved in synthesis, structural characterization, and investigations of physical properties of perovskite compounds in the form of three-dimensional (3D) bulk crystals, two-dimensional (2D) nanosheets, one-dimensional (1D) nanorods or nanowires, and zero-dimensional (0D) quantum dots (QDs) or nanocrystals [3]. Perovskite materials exhibit fascinating and unique physical properties that have been extensively studied for both practical applications and theoretical modeling [4]. In this sense, perovskite materials’ potential applications are varied and include uses in sensors, fuel cells, solar cells, photodetectors (PDs), memory devices, lasers, and spintronic applications [5,6]. Among the above, PDs are an essential optoelectronic component found to have a wide range of applications, both in industry and daily life, including astronomy, surveillance, robotics, smartphones, and environmental monitoring [7,8]. Until recent years, PDs have been made up of inorganic semiconductors such as Si, InGaAs, and GaN, which are used to detect light in the visible range (450–800 nm), infrared (900–1700 nm) range, and UV range (250–400 nm), respectively [4]. Despite the fact that these PD technologies possess mature and authentic fabrication processes, their widespread application and commercialization are impeded by complex and expensive manufacturing processes, mechanical inflexibility to current smart systems, and the requirement of high driving voltage. For example, the photoresponsivities of commercially available Si-, Ge-, and InGaAs-based PDs are usually around few A/W under high bias voltages (5–200 V). Therefore, an alternative is needed to overcome the existing problems; searching for novel material is a motivation.

Meanwhile, in the near future, the era of the Internet of Things (IoT) will integrate sensors and objects with networks and solely play an eminent role in world economic development [9]. With the predictable trending, the smart sensor network, as an inevitable component of the IoT, will become a key field in deciding the future development of information technology [10]. Understandably, the smart sensor network requires a great amount of electric energy for sustainable and maintenance-free operation. However, such a huge power cannot be provided due to the huge number of sensor networks and the complexity of replacing batteries every time. Therefore, wireless devices should be self-powered without using batteries.

In general, a conventional PD needs to be operated by an external power source; typically, it is the battery. However, such independent power supplies are not compatible with a future intelligent sensor system in the following ways: (1) the material used for the battery construction is likely to be highly hazardous to the ecosystem; and (2) the requirement of battery recycling has to be considered in terms of cost for an integrated PD network. Therefore, independent, sustainable, and maintenance-free PDs can be operated by an built-in power source or by extracting power from the surrounding environment. So, a PD operated under self-biasing mode is called a self-powered PD.

In recent years, self-powered electronic devices (in other words, PDs) have received extensive attention along with the rapid development of smart systems and wearable electronics in daily life, for example, smart homes, bioimaging, health monitoring, and optical communications [11]. Generally, self-powered PDs are classified as Schottky diodes, p-n junction diodes, and metal–semiconductor–metal (MSM) diodes based on device structure and junction formation [7,11–16]. By now, there have been numerous reports on SPPDs, which can sense from the deep ultraviolet (DUV) to far infrared regions (FIRs). Though the performance metrics are very adequate, currently all the proposed self-powered perovskite photodetectors (SPPDs) are not suitable for commercial applications due to the complexity of device fabrication, the cost factor, and the mechanical rigidity, which make them difficult to integrate with electronic components. With the benefit of longer carrier diffusion length and high charge carrier mobility, high absorption coefficient (α≈105cm−1), and high defect tolerance, perovskite materials could be the best choice for self-powered PDs. These excellent optoelectronic properties allow detection of light to be realized within the perovskite layers, even with 100–200 nm thin films, which is ideal for high-resolution imaging applications, whereas conventional technologies such as silicon-based imaging devices require micrometer-range thick layers. Furthermore, the solution processed manufacturing and the ease of fabrication process of perovskite layers at low temperatures compared with conventional complementary metal–oxide–semiconductor (CMOS) technology make it attractive for future development of electronics field. Therefore, these additional advantages of perovskite materials are potentially attractive for large-area manufacturing at low-cost production, which provides fabrication of self-powered PDs on flexible polymer substrates.

There have been a considerable number of review papers that separately covered perovskite-based PDs, nanoscale self-powered PDs, and self-powered UV PDs [17–19]. However, a comprehensive review focused mainly on self-powered perovskite-based-PDs, specifically on fabrication technique, device performance in-terms of photoresponsivity, specific detectivity, response speed is lacking in the literature. This review is mainly focused on self-powered perovskite-based PDs. First, we introduce the working principle and basic mechanism of self-powered PD systems in two types of modes: the photovoltaic (PV) mode and the integrated-self powered system. Then, we summarize the recent progress on self-powered perovskite-based PDs by sectioning the structure of perovskites such as bulk crystal structure (3D), nanosheets (2D), nanowires or microwires (1D), and QDs or nanocrystals (0D). Besides, we introduce flexible self-powered perovskite photodetectors (SPPDs). Despite the considerable advancement in the field, there are several key challenges ahead to face in the further development of SPPDs. Therefore, we have provided a survey of challenges and opportunities in the last section.

Despite the remarkable device performance offered, organic-inorganic perovskites suffer from degradation by air, moisture, temperature, electric field, and light exposure (detailed analysis is provided in challenges and perspectives section later) [30,31]. Particularly, the organic cation clusters [CH3NH3+, FA+, CH(NH2)2+, etc.] in the organic-inorganic hybrid lead halide perovskite are extremely sensitive to the environmental humidity and oxygen content, leading to problems such as poor environmental stability and poor thermal stability [32,33]. Furthermore, the perovskite material intrinsically suffers from issues such as migration of ions and the appearance of hysteresis in the electrical characteristics [34,35]. The making of stable functional devices is greatly restricted by these issues. Therefore, a stable perovskite requires a partial replacement organic counterpart with more stable cations such as long-chain organic cations or inorganic elements such as cesium (Cs) and rubidium (Rb) [36–38]. A representative all-inorganic perovskite CsPbBr3 is a direct band-gap semiconductor with low trap state density, high carrier mobility, long electron-hole diffusion length, etc., and it also has some excellent optical characteristics such as high quantum yield, strong light absorption, high luminous efficiency, and adjustable luminous wavelength (see Table 1) [1].

Here, some very important performance parameters to describe the characteristics of a PD can be summarized as follows.

Responsivity: The responsivity is given as the ratio of the output current or voltage to the power of the input light signal, and the unit is A/W or V/W. R=IpPA,(1)where Ip=Ilight−Idark, P is the incident power density, and A is the effective device area.

Specific detectivity: D* is a key figure of merit to reflect the sensitivity of a device. In general, the specific detectivity of the PD is related to the noise inversely. It is defined in terms of responsivity R as D*=(AΔf)1/2inR=RA2qId,(2)where R is the responsivity in A W−1, A is the effective device area, Id is the dark current, and q is the coulombic charge of the electron (1.6×10−19C).

External quantum efficiency (EQE): EQE is the ratio of the output number of electrons/holes to the number of incident photons, indicating the conversion efficiency of PDs from photons to charges. It is given as follows: EQE=Rhcqλ,(3)where h, c, q, and λ denote Planck’s constant, the speed of light, elementary electron charge, and wavelength, respectively.

Noise equivalent power (NEP): NEP is the signal power that produces a signal-to-noise ratio (SNR) to be equal to 1, representing the minimum impinging optical power that a PD can distinguish from the noise. It can be written as NEP=(AΔf)1/2D*=inR,(4)where A is the effective device area, Δf is the electrical bandwidth, and in is the noise current.

Linear dynamic range (LDR): It describes a light intensity range in which the current response of the PD is linearly proportional to the light intensity, and it can be calculated by LDR=20logJmax−JdJmin−Jd,(5)where Jmax is the upper photocurrent, Jmin is the lower photocurrent, and Jd is the dark current. Detecting both the weak and strong light is required for a high LDR PD.

Response speed: Rise time (tr) and decay time (td) are defined as the time within which photocurrent increases from 10% to 90% and drops from 90% to 10% of its maximum value.

Energy consumption is one of the important aspects considered for modern electronic devices which needs further development to achieve a better sustainable future. This is equally true for commercially available PDs. In recent years, there have been numerous reports on SPPDs operated from the DUV to the near-infrared (NIR) under zero bias voltage. The SPPDs are categorized into two types based on energy source feeding as follows: (1) PV effect-based PDs (Schottky junction and heterojunction) and (2) PDs with integrated self-powered systems.

Apart from SPPDs based on the p-n heterojunction, PDs based on the Schottky junction can also operate without external power sources owing to the PV effect. Furthermore, compared to p-n heterojunction PDs, Schottky-type PDs exhibit ultrarapid response time, high photosensitivity, and low-cost fabrication, which is highly preferable for future applications. Generally, a built-in electric field, which separates the photogenerated electron–hole pairs and gives rise to circuit current, is formed due to the electron’s spontaneous diffusion caused by the work function difference between contact metal and semiconductor. Unfortunately, the surface state of semiconductors could seriously affect the diffusion process. Therefore, to achieve high performance, a great endeavor is required to optimize the stability and quality of the Schottky contact. Up to the present, several investigations of self-powered Schottky-type PDs made from different semiconductors have been reported [18,40–45].

For the realization of self-powered PDs in PV mode, the following aspects should be considered: (1) SPPDs can not only detect the signals but also must be powered by the detected signals; (2) the photogenerated electron–hole pairs are usually separated by the built-in potential difference provided by the junction-based multilayer structures, which often involve complicated, time-consuming, and uneconomic device fabrication processes [46]; (3) more importantly, the material choice of this kind of device is limited due to issues such as lattice mismatch, surface states, and band alignment [47–49]. These issues not only increase the system size but also greatly limit mobility and independence [50]. In this regard, a miniaturized, uninterruptible energy source is necessary to power up the PD.

The nanogenerator (NG) is a new technique, first proposed by Wang in 2007, that utilizes mechanical and thermal energies produced by human body motion and then converts into electrical energy [51]. Generally, NGs can be classified into three types based on electricity generation modes: the triboelectric generator (TENG), the piezoelectric NG (PENG), and the pyroelectric NG (PYENG). A TENG is an energy-harvesting device that converts mechanical energy into electrical energy by a combination of triboelectric effect and electrostatic induction. A PENG is a device capable of converting external kinetic energy into electrical energy via motion by piezoelectric materials. The conversion of external thermal energy into electrical energy is adopted for designing PYENGs. Among the above, TENG is a compatible nanoenergy source that is frequently used to back up the electronic devices and has drawn more attention. These NGs are widely used as the micronanoenergy sources for self-powered sensors. However, integration of NGs with sensor devices is always challenging, but further development of NG-based self-powered sensors is extremely attractive.

The single-crystal perovskites possess many unique advantages over polycrystalline thin-film structures, such as high purity, fewer grain boundaries, and enhanced thermal and moisture stabilities. Notably, in a single crystal, low trap density contributes to high carrier mobility and long carrier diffusion lengths, resulting in highly sensitive PDs. High-purity perovskite single crystals have been prepared by several methods reported earlier such as inverse temperature crystallization (ITC) [52], antisolvent vapor-assisted crystallization [26], top-seed solution growth [53], bottom-seeded solution growth [54], and solvent acidolysis crystallization [55].

Similarly, Zhang et al. have reported on a vertical 2D/3D perovskite heterostructure (4-AMP) (MA)2Pb3Br10/MAPbBr3 [4-AMP-4-(aminomethyl) piperidinium], featuring a well-defined interface and high crystalline quality [57]. The photographic image of single-crystal heterostructure and corresponding cross-sectional field emission scanning electron microscopy (FESEM) image are shown in Fig. 3(d). Electrical transport measurements demonstrate that the (4-AMP) (MA)2Pb3Br10/MAPbBr3 heterostructure can form a vertical diode with obvious current rectification behavior and photocurrent generation characteristics. Benefiting from the built-in electrical field at the junction [Fig. 3(e)], PDs based on those millimeter-thickness heterostructure crystals exhibit high performance in self-driven operation mode, including fast response time (600/600 μs), and high detectivity (≈1012Jones) [Figs. 3(f) and 3(g)].

In the CH3NH3PbI3 crystal, a [PbI6] octahedral structure is constituted by Pb and I atoms, and CH3NH3+ is located in an octahedron cavity. The distortion of the [PbI6] octahedron results in the Pb2+ not occupying the center of the octahedron, and so the crystal symmetry is destroyed, inducing a spontaneous polarization in the perovskite. The spontaneous polarization of the CH3NH3PbI3 unit cell and the ordered arrangement of the unit cell in the single crystal provide the possibility of the single crystal achieving self-powered PDs. For instance, a self-powered PD in a MAPbI3 single crystal was reported by Zhang et al., where the symmetric electrode is fabricated in Au/MAPbI3/Au [60]. However, from the device structure of Au/MAPbI3/Au, if the MAPbI3 crystal is nonpolar, two Schottky junctions with the same barrier height will be formed at both ends of the device. Such a device has no built-in electric field and cannot realize the self-powered characteristic, which is contrary to the experimentally observed self-powered phenomenon. Therefore, the ferroelectric spontaneous polarization exists in the MAPbI3 single crystal, which is responsible for the self-powered mechanism of PDs. Briefly, the photogenerated carriers in ferroelectric materials are separated by the depolarization field associated with spontaneous polarization and collected by the Au electrodes at both ends of the device to generate the potential difference. The process of carrier movement is shown in Fig. 4(d). Moreover, Fig. 4(e) shows that the photoresponsivity of the device greatly depends on the channel size, i.e., increased channel size could effectively increase the illumination area of the device and enhance the photocurrent, while the carrier collection efficiency is simultaneously reduced, and then the responsivity of the device is decreased. Without external bias, the device exhibited the maximum responsivity of 0.16AW−1 and the specific detectivity of 5.86×1011 Jones with a fast response time.

Recently, low-dimensional nanostructures have been extensively studied as a potential building block to construct efficient PDs. Compared with bulk materials, low-dimensional nanoscale materials, with their large surface areas and possible quantum confinement effect, exhibit distinct electronic, optical, chemical, and thermal properties [61]. One should consider that poor structural stability and chemical stability remain major concerns for the practical application of halide perovskites. So, the high surface-area-to-volume ratio of nanostructured perovskite can increase the impact of surface properties on the chemical properties and phase stability. However, nanostructures of halide perovskites can exhibit enhanced structural and chemical stability owing to a surface energy effect and surface ligand functionalization [62]. Therefore, a large variety of perovskite nanostructures, such as QDs/nanocrystals, nanowires/nanorods, and nanosheets, were successfully synthesized, which could be effectively applicable in PDs [3,63].

Apart from the Schottky junction PD, perovskite QDs are used as a photosensitive layer in multilayer junctions to realize heterojunction-based self-powered PDs. For example, Imran et al. proposed a trilayer PD with device structure ITO/ZnO/CdS/CsPbBr3/Au [67]. To reduce the interfacial charge carriers’ recombination and the charge transport resistance, CdS nanorods are sandwiched between a ZnO/CsPbBr3 interface. The cross-sectional FESEM image of the device is shown in Fig. 5(f). The typical I-V curve of the device represents the PV effect by the formation of an open-circuit voltage of 0.13 V, as shown in Fig. 5(g). The device performance could be explained based on the band diagram, as illustrated in Fig. 5(h). In the trilayer device, built-in potentials are induced at both the CdS/CsPbBr3 and ZnO/CdS interfaces, and both CsPbBr3 and CdS nanorod layers can absorb the incident light. This doubly formed built-in potential in the trilayer device results in more efficient separation of photogenerated carriers and then more efficient drift to electrodes at zero bias. Overall, the self-powered trilayer PD shows photoresponsivity of 86mAW−1 and specific detectivity of 6.2×1011 Jones with the rise and decay time of 0.3 s and 0.25 s, respectively.

Similar to 0D perovskite QDs, 1D perovskite materials also found potential application in PDs owing to their high sensitivity, reduced recombination rate, and quick charge transfer characteristics. Until now, a great number of the studies carried out by researchers on the fabrication of PDs used various perovskite nanowires/nanorods, microwires, microtubes, etc. [68–73]. However, nanowire/nanorod-based self-powered PDs have seldom been reported in the literature. The reasons can be summarized as follows: (i) complicated fabrication process of nanowires (NWs) as a device structure and poor reproducibility; (ii) the photocurrents of the PDs based on aligned NWs or a single NW with MSM structure are very low due to the limited light absorption cross section or the large channel length between the metal electrodes; (iii) a phenomenon of the p-n junction and the Schottky junction required for achieving self-powered PDs is very complicated in individual NW PDs or aligned NW PDs due to difficulties of p or n doping in perovskites and tedious nanowire manipulation [68,74].

The stability of perovskite PDs, as well as the selection of photoactive material, is the most important issue taken into consideration while designing the structure of PDs. Li et al. demonstrated the fabrication of CsxDMA1−xPbI3 thin films by the solution-processing technique [89]. Instead of PbI2, DMAPbI3 [DMA+=dimethylammonium, (CH3)2NH2+] is chosen to react with CsI, which can enhance the room-temperature phase stability of resultant Cs-based perovskite films. A self-powered PD has been designed, as indicated in Fig. 8(l), which shows broad photoresponse from 300 to 750 nm, with high specific detectivity (>1×1013 Jones) [Fig. 8(m)]. Above all, the device shows excellent long-term stability in the air due to an increased tolerance factor by the introduction of DMA+; in other words, DMA+ prevents the spontaneous transformation of CsPbI3 at room temperature [Fig. 8(n)].

In perovskite-based optoelectronic devices, ion migration is one of the interesting concepts to be investigated to understand unusual phenomena such as current-voltage hysteresis, switchable PV effect, and light-induced self-poling effect [35,91,92]. Pang and coworkers investigated the self-poling effect in perovskite thin films by designing the PD in a metal–oxide–semiconductor (MOS) structure [90]. The device with structure Si/SiO2/CH3NH3PbI3 showed excellent self-powered photoresponse characteristics and exhibited an excellent on/off ratio of 105 with a bias of 0 V and a fast response time of 25.8 ms. The authors explained the “self-biasing” by the concept of self-poling effect as follows. When no bias is applied, the mobile ions within the perovskite layer show electrical disorder in the dark, and the perovskite has a uniform Fermi level, as shown in Fig. 8(o) (upper side). Under illumination, the built-in electric field is formed when the positive ions/vacancies accumulate at the perovskite/SiO2 interface and the negative ions/vacancies accumulate at the perovskite/Pt interface due to the preset built-in potential caused by band bending in the asymmetrical device [Fig. 8(o) (lower side)]. The typical band diagram elucidated from XPS spectra is shown in Fig. 8(p) and clearly explains the charge transport properties across the device.

The ultra-flexible, ultrafast (<20ms), wearable, and flexible PDs are the highly studied research field due to their compatibility with a variety of emerging areas such as flexible, stretchable, wearable, portable, and printed optoelectronics. Combined with transparent features, flexible devices can be employed in touch screens and interactive electronics [114]. There are so many novel functional materials including QDs, inorganic nanostructures, 2D layered semiconductors, organic semiconductors, etc., which have been extensively studied as the active layers in flexible optoelectronic devices. Particularly, metal halide perovskites exhibit excellent electrical and optical properties, as well as outstanding mechanical stability, and they have been used as cost-effective flexible PDs [8,115,116]. For high-sensitivity flexible PDs, both high photoresponsivity and mechanical flexibility should be implemented simultaneously in a single device, which sets a great challenge for fabrication techniques and selection of materials [117]. High-temperature processing is also another important factor in order to improve the crystallinity of photosensing material in some cases; there are great obstacles to fabricating devices on flexible substrates such as plastic. Cost-effective perovskite material, which exhibits advantages of solution processing and low-temperature fabrication is the promising light-harvesting material for fabrication of high-performance flexible PDs. The majority of flexible devices fabricated to date have been on ITO transparent conducting electrodes and are fabricated by a solution-processing technique at low temperatures, but ITO is also not the best choice owing to its fragility. Therefore, some other alternatives have also been used, such as carbon nanotubes, graphene, metal nanowires, and conducting polymers [118–122]. Overall, flexible PDs must be fabricated at low temperatures and have specific advantages such as a simple processing technique, low cost, shock resistance, light weight, durability, and portability.

In another approach, the appropriate doping of the photoactive layer would enhance the PD performance and ensure flexible compatibility. Toward that direction, Shin et al. reported the photodiode structure using graphene QDs (GQDs): CH3NH3PbI3 as the photoactive layer and PEDOT:GQDs as the hole transport layer (HTL) [124]. By incorporating GQDs, along with the increase of perovskite layer crystallinity, the work function of PEDOT: GQD HTL is also increased due to modification of PEDOT from benzodide to quinoid by adding carbon-based material, which leads to the efficient generation of charge carriers and reduced recombination at the interface. The typical device structure and its corresponding cross-sectional scanning electron microscopy (SEM) image are presented in Fig. 11(e). The consecutively enhanced photoresponsivity of 0.354AW−1 and specific detectivity D* of 8.42×1012 Jones were observed, as shown in Fig. 11(f). Furthermore, the flexible PDs also show excellent stability by maintaining 80% of the initial responsivity even after repeated bending for 1000 cycles at a bending radius of 4 mm [Fig. 11(g)].

Instead of mechanically actuated NGs, the in-built solar cell system can also be an excellent choice for SPPDs. For example, an all-perovskite self-powered nanosystem was demonstrated by Li et al., who assembled the perovskite solar cell with a perovskite PD, as shown in Fig. 13(d) [130]. The authors fabricated planar configuration perovskite solar cell using TiO2 as an electron transport layer, CH3NH3PbI3 as the light-absorbing layer, and P3HT as the hole transport layer. The as-fabricated solar cell was connected to ITO/CH3NH3PbI3/ITO configured planar PD on a flexible PET substrate through Cu wire, and another side of the PD was connected with the amplifier to measure photocurrent. As an energy conversion unit in the nanosystem, the perovskite solar cell with a high efficiency of 10.5% drives the light-sensing unit to achieve the detection of various lights. Under AM1.5 irradiation (100mWcm−2), the perovskite solar cell can provide a 0.93 V voltage for the PD [Fig. 13(e)]. Furthermore, the mechanical stability and durability of the PD were examined by multiple bending exercises and no obvious degradation of photoresponse behavior was observed, even after 200 bending cycles, implying good mechanical endurance, as shown in Fig. 13(f).

In addition to the aforementioned SFPD structures, numerous reports have appeared in the literature, as summarized in Table 3. As an example, nanowire array-based PDs show poor photoresponse performance due to the existence of multiple microinterfaces between randomly oriented nanowires, which has been systematically studied by Zeng et al. [131]. The authors fabricated nanowire network (NWN) PDs with a welding strategy that showed ultrahigh performance with an on–off ratio and detectivity of 2.8×104 and 4.16×1013 Jones, respectively. More importantly, the unpackaged NWN PDs show ultrahigh storage stability in the air with a humidity of 55%–65%, and the flexible NWN PDs can withstand 250 bending cycles at different bending radii and 1000 bending cycles at fixed bending radii with no performance degradation being observed. Also, a self-powered flexible fiber-shaped PD based on double-twisted perovskite-TiO2-carbon fiber and CuO-Cu2O-Cu wire was designed by Li et al. and achieved an ultrahigh detectivity of 2.15×1013 Jones and response time of less than 200 ms [136]. Recently, Ogale et al. demonstrated the flexible self-powered PD by forming a heterojunction of SnO3 and cubic phase α-CsPbI3 [101]. The cubic phase of CsPbI3 was stabilized by polyvinylpyrrolidone (PVP) wrapping, by which the device performance and environmental stability of the device have been notably enhanced.Table 3.

Summary of Flexible Self-Powered Perovskite-Based PDs

The perovskite-based SPPD operable at various subbands from DUV to the NIR has been achieved. However, the significant advantages and disadvantages regarding perovskite-based PDs are applicable to perovskite-based self-powered PDs too. Despite significant development in perovskite PDs, there are still formidable issues and challenges to be resolved to shift from laboratory to industrial mass production and application. The performance can be further improved by optimizing the intrinsic properties of perovskite material and device fabrication schemes. Apart from key parameters such as photosensitivity, photoresponsivity, detectivity, and response speed, the important criteria for practical device application are that the device has to maintain a stable photocurrent and dark current for the long term under standard conditions. This criterion has mainly been affected by issues such as moisture, thermal condition, and photoinstability [137,138].

Mostly, perovskite-based optoelectronic devices are based on hybrid organic-inorganic perovskite materials, which have noncoordinated ions such as Pb2+, I−, and MA+ that were responsible for the perovskites to be highly sensitive to moisture and polar solvent and subsequently affect the stability of the perovskite [139,140]. It was found that large-sized cations (i.e., long-chain organic cations) can enhance the stability of perovskite materials [89,141]. The perovskite materials with long-chain cations showed negligible degradation after exposure to moisture for 46 days. It was also reported that the Pb-containing perovskites are stabler in air than their Sn-containing counterparts because Sn2+ is easy to be oxidized to Sn4+. Specifically, the Pb-containing perovskites show only surface degradation in the dark with the bulk properties of the materials retained for a couple of weeks in air. Moreover, it was observed that the metal luster in MAPbI3 single crystals can even keep for more than half a year in air. In contrast, the Sn-containing materials are air- and moisture-sensitive and partially decompose within 2 h before total decomposition after 1 day [142]. Apart from the crystal structure of perovskite, the defect structure of perovskites plays an important role in the deterioration of device performance [143]. The presence of defects in light-sensitive material has commonly been recognized as fatal, such that charge mobility, carrier lifetime, and conductivity are decreased, which are inevitable properties for optoelectronic devices. A stability study on polycrystalline thin film performed by Wang et al. reported that deterioration of perovskite occurs due to the large number of grain boundaries and surface defects. On the other hand, single-crystal perovskites with no grain boundary and diminished surface defects show better stability in the air for a longer time [144]. It has been concluded that the hybrid perovskites tend to decompose due to the hygroscopic nature of amine salts, and degeneration mainly starts at the structural defects of crystal as well as at grain boundaries [145].

Perovskite semiconductors demand deep investigation with respect to their stability in terms of moisture and temperature. However, the encapsulation technique could give the solution to enhance the stability to some extent. Stability under light exposure is another challenge to be faced while dealing with perovskite-based PDs [137]. The deep insight into the mechanism of light-induced degradation will be valuable for designing highly stable perovskite materials [147]. There are many reports on the degradation of perovskites under light illumination. The hypotheses include electronic trap states [148], photochemical reactions [149,150], activate transport of halide ions and/or organic cations [151], etc. Bag et al. studied the stability of perovskite under light exposure by replacing an MA+ (methylammonium) ion with an FA+ (formamidinium) ion; stability was enhanced by optimizing the balance between MA+ and FA+ ions [151]. Hence, it can be concluded that the controlling of vacancy defects and ion migration deficiency can considerably enhance the stability of perovskite-based devices under long-term exposure to light illumination.

Several issues and open questions regarding the commercialization of perovskite-based PDs remain to be confronted. The toxicity caused by lead is always a problem, threatening human health, poisoning organisms, polluting the environment, and causing difficulty in discharging lead from the body. Research indicates that the contamination of lead ions to soil and water sources is permanent and generates a very serious negative impact on human, animal, and plant survival [152–155]. Therefore, to assure human safety and a pollution-free natural environment, it is very essential to develop some non- or low-toxic metal ions to replace lead as perovskite materials. For example, there have been multiple previous attempts to replace Pb2+ ions by less toxic ions such as Sn2+, Bi3+, Ge2+, Sb3+, Mn2+, and Cu2+ ions [156–160]. These efforts not only increase the variety of perovskite materials but also enhance the environmentally friendly features. Although the toxicity has been weakened by lead-free perovskites, the performance of the device needs to be further improved. Among the perovskite components, lead acts as a core skeleton that plays an important role in the generation and transport of charges. The carriers transfer along with the Pb–halogen bond, resulting in poor performance and lower stability after replacing Pb with Sn or Bi. Many researchers have proved that the role of lead is irreplaceable; thus the toxicity treatment of perovskite needs to be further considered. Overall, to facilitate a profound understanding of perovskite characteristics and device physics, the focus should be sharpened to their crystal growth process, defect engineering, device fabrication technique, and more importantly, the stability of devices under ambient conditions.

In summary, by virtue of their superior optoelectronic properties, perovskite materials have made a giant step in the PDs research field. Although significant advances in the development of perovskite-based self-powered PDs have been made in past years, there are still some challenges remaining before moving forward with practical applications. There are two ways to bring the perovskite-based PDs to practical or commercial application: either by optimization of material synthesis with high crystal quality and enhancement of stability issues in perovskite material or advancement in device fabrication strategy. Moreover, PD arrays are less explored for real applications in imaging and biomedical sensing, which should be a focus in the future. Specifically, the fabricated self-powered perovskite PDs should be intelligent, multifunctional, supersmall, extremely sensitive, and energy-efficient. This requires the rational synthesis of materials, fabrication of devices, and integration of various devices into a system with multifunctional characteristics and operation without external power sources. We strongly believe that the reader can acquire more comprehensive knowledge in this field while reading this review and motivate young researchers to undertake the tasks to solve the issues raised in this review.