Ultraviolet Photocatalytic Degradation of Perovskite Solar Cells: Progress, Challenges, and Strategies

The certified efficiency of perovskite solar cells (PSCs) has reached 25.5% within just around 10 years, approaching the highest reported value of mainstream silicon solar cells. The application of metal–oxide (MO) electron transport layers (ETLs), including TiO2 and SnO2, is crucial for achieving highly efficient PSCs because of their wonderful photoelectrical properties, resulting in n–i–p conventional structure devices, constantly breaking the world record efficiency. However, these MOETLs inevitably lead to the degradation of PSCs due to their photocatalytic activity under actual sunlight which includes ultraviolet (UV) range. Overcoming the UV photocatalytic degradation is still a great challenge for the state‐of‐the‐art PSCs toward practical applications. Herein, the recent progress related to the UV photocatalytic degradation of PSCs induced by the MOETLs based on recent literature reports is reviewed, including the photocatalysis origin, degradation mechanism, challenges, and various strategies. Perspectives for future efforts in overcoming UV photocatalytic degradation are provided. It is believed that this review is advantageous for achieving stable PSCs under actual outdoor sunlight.


Introduction
Perovskite solar cells (PSCs) have received great attention due to their ever-increasing power conversion efficiency (PCE), low-cost materials, and easy solution preparation. The certified efficiency of PSCs reached 25.5% [1] based on a lab scale, exceeding the performance of copper indium gallium selenide (CIGS) and CdTe thin-film solar cells and approaching the highest reported value of the mainstream silicon solar cell. [2] The highest-efficiency devices generally use the regular n-i-p structure composed of a perovskite absorber between a metal oxide electron transport layer (MOETL) and an organic hole transport layer (HTL). [1,3] Miyasaka et al. first adopted CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 as light absorbers in liquid electrolyte-based dyesensitized solar cells using a thick TiO 2 layer (%10 μm) as an electron collector. The first PSCs achieved a PCE of 3.8%. [4] After that, tremendous efforts have been made for the PSCs based on TiO 2 . The PSCs included an n-type TiO 2 ETL, a perovskite absorber layer, a p-type 2,2 0 ,7,7 0tetrakis-(N,N-di-4-methoxyphenylamino)-9,9 0 -spirobifluorene (Spiro-OMeTAD) HTL, which has been the most typical and successful structure. Grätzel et al. achieved the first certified efficiency of 14.14% based on the aforementioned n-i-p structure with compact TiO 2 (c-TiO 2 )/mesoporous TiO 2 (m-TiO 2 ). [5] In the following years, the performance of the PSCs based on TiO 2 achieved tremendous progress. To remove the organic material in TiO 2 paste and achieve high-quality m-TiO 2 layer, a hightemperature sintering process (>500 C) was generally needed. [6] To avoid high-temperature processing, low-temperature processed SnO 2 was developed. Tian and co-workers adopted SnO 2 thin films attained by spin coating SnO 2 nanoparticles on indium-doped tin oxide (ITO) followed by low-temperature annealing at 200 C, and a 13% PCE was achieved. [7] Huge progress of SnO 2 -based PSCs was reported by Fang and co-workers. They adopted thermal decomposition of SnCl 2 ·2H 2 O precursor at 180 C in ambient air to form SnO 2 film and achieved 17.21% efficiency. [8] Later on, Hagfeldt et al. used low-temperature chemical bath deposition to grow SnO 2 ETL and obtained almost hysteresis-free PCEs of 20.8%. [9] In 2019, the PCE of the planar SnO 2 PSCs reached a certified efficiency of 23.32%, which first surpassed that of m-TiO 2 PSCs. [10] Therefore, SnO 2 has increasingly attracted great attention as an ETL for PSCs, and it is considered as the most promising alternative to TiO 2 . Up to date, all the world records of the certified PCE were achieved based on TiO 2 or SnO 2 n-i-p conventional structures since the PSCs appeared in 2009 ( Figure 1). [1,3,5,6,[10][11][12][13][14][15] Recently, a PCE of 25.8% (certified 25.5%) has been achieved based on Cl-bonded SnO 2 . [1] Thus, the important key to achieve state-of-the-art PSCs is the application of TiO 2 or SnO 2 , due to its excellent electrical properties.
However, it is known that the excellent photoelectric properties of TiO 2 and SnO 2 make them to be "golden" photocatalysts simultaneously. In recent decades, TiO 2 and SnO 2 photocatalysis have been widely studied in various areas, such as water splitting and pollutant degradation. [16][17][18][19] When TiO 2 and SnO 2 are exposed to near-ultraviolet (UV) light, it will inevitably decompose the adjacent perovskite absorption layer and thus decrease the stability of PSCs. [19,20] In the International Electrotechnical Commission (IEC) standards on terrestrial photovoltaic (PV) modules, light stability requirements include 1) UV preconditioning test: the module should be exposed to high-intensity and continuous UV irradiation with a total exposure of 15 kWh m À2 and 2) light soaking: the module is to be irradiated with a total of 86 kWh m À2 of continuous standard outdoor sunlight (including UV). [21][22][23] Such instability cannot be avoided during device operation, even if they are well encapsulated. It is worthy to note that most of the reported top-level photostabilities of PSCs are always carried out by completely eliminating the UV light using a UV filter or adopting a white light-emitting diode (LED) without UV range. [24,25] It illustrates the importance of suppressing the photocatalytic activity of TiO 2 and SnO 2 for stable operation of PSCs.
In this review, we will focus on the recent progress related with the UV photocatalytic degradation of PSCs, including degradation mechanism, challenges, and strategies. First, we briefly summarize the recent progress and trends of the state-of-the-art PSCs and emphasize the importance of overcoming the UV photocatalytic degradation. Next, we review the degradation mechanism of perovskites under the UV photocatalytic effect of TiO 2 and SnO 2 . Then, we provide a comprehensive summary of the recent efforts of improving the PSC stability against the UV photocatalytic effect through interface engineering, removing the UV, using ETLs with inferior or no photocatalytic activity, etc. Finally, after assessing these methods, we provide our perspective and views of strategies that suppress and even eliminate the UV photocatalytic degradation effect in efficient PSCs. We hope that this review will be advantageous to improve the operational stability of efficient n-i-p PSCs under outdoor natural light soaking without cutting the UV light.

The Properties of TiO 2 and SnO 2 as ETLs in PSCs
As mentioned earlier, the wonderful photoelectrical properties of TiO 2 and SnO 2 are crucial to achieving high efficiency in n-i-p conventional structure PSCs; however, their UV photocatalytic activity largely harms device operational stability.

Excellent Photoelectric Properties
As ideal electron transport materials for PSCs, they need to meet some requirements: 1) well-matched energy alignment to trigger electron transfer while blocking holes; 2) high transparency to allow efficient light harvesting; 3) excellent electron mobility to minimize charge accumulation; and 4) easy access to starting  [4,[7][8][9] and red champion flags [1,3,5,6,[10][11][12][13][14][15] show the important milestones and the certified PSCs' performance given in NREL. materials with high chemical resistance to perovskite solvents ( Figure 2a). [26,27] TiO 2 and SnO 2 seem to be almost perfect ETLs for application in PSCs as they meet the earlier characteristics well. [28,29] 2.1.1. Band Structure TiO 2 mainly occurs in nature as the following well-known minerals, rutile, anatase, and brookite, which have different characteristics of Ti-O bonds. [30] Among these three phases, the anatase phase has a wider bandgap of 3.2-3.5 eV and it is the most common one in PSCs. The bandgaps of SnO 2 reported range from 3.5 to 4.0 eV, depending on its specific synthesis conditions. The suitable conduction band (CB) of TiO 2 and SnO 2 (3.8-4.3 eV) contributes to the effective electron extraction from perovskite and the deep valence band (VB) around 7.2-7.9 eV can effectively prevent the hole from reaching the transparent electrodes. [31,32] With a wide bandgap, they are advantageous in maximizing the light absorption of the perovskite layer by allowing more light to pass through them. The good band alignment of TiO 2 or SnO 2 with the perovskite layer theoretically ensures a high quality of the p-n heterojunction, which implies improved electric parameters.

Optical and Electrical Properties
TiO 2 and SnO 2 have superior light transmittance due to the wide bandgap and small reflective index of 2.4-2.5 and <2, respectively. [33] The efficient light management in the UV-visible region offers a photon which can pass through easily and be absorbed by the perovskite absorber. The electron mobility of bulk TiO 2 is about <1 cm 2 V À1 s À1 , and there are many effective ways to enhance the electron transfer capability, such as metal doping. [34,35] In case of SnO 2 , it has high electron mobility (up to 421 cm 2 V À1 s À1 ) and high conductivity (%1.7 S cm À1 for crystalline and %7.2 Â 10 À3 S cm À1 for amorphous SnO 2 ), suggesting high electron collection efficiency and transport ability. [36]

Chemical Stability and Easy Preparation
The TiO 2 and SnO 2 materials have excellent chemical stability. The lower hygroscopicity and acidity resistance contribute to the durability of the PSC device, which enables compatibility with various fabrication methods. [37] The explored techniques for the preparation of TiO 2 and SnO 2 have been reported with several different methods, mainly including the sol-gel method, nanocrystal approach, chemical bath deposition, the atomic layer deposition technique, electrodeposition, and so on. [32,33] 2.2. Photocatalytic Activities of SnO 2 and TiO 2 As the bandgap of TiO 2 and SnO 2 (3.2-3.8 eV) corresponds to 410 and 326 nm, the excitation wavelength almost falls in the UV region. Given that %3-5% of UV radiation in total solar flux is incident at the Earth's surface, [38,39] the excitation of MO by photons with light energy greater than the bandgap will generate holes and electrons, leading to interfacial redox reactions named the photocatalytic reaction ( Figure 2b). [18,40] For MO photocatalysis, the generation of the electron-hole pair can be written as follows: MO þ hv ! e À (MO) þ h þ (MO); electrons in the filled VB will be excited to the vacant CB, leaving holes in the VB. After electronhole pair separation, the electrons or holes that migrate to the surface will drive the reduction or oxidation reactions, respectively. Therefore, a whole photocatalytic reaction can be broken down into the following two half reactions: an electron-induced reduction reaction and a hole-induced oxidation reaction. Photocatalytic reactions generally occur at the MO surface. It means that the photocatalytic process will inevitably occur in PSCs during their operation under natural solar light. However, the characterization of these fundamental processes is very complicated.

Photocatalytic Degradation Mechanism of PSCs
Although the UV photocatalytic degradation of PSCs is widely mentioned in numerous reports, studies related to the mechanism for the degradation reaction of perovskite are still rare. Figure 2. a) The typical n-i-p device structure and the requirements of ETL and b) the schematic processes in MO photocatalysis. Reproduced with permission. [40] Copyright 1995, American Chemical Society.

Photocatalytic Degradation of PSCs Induced by TiO 2
Snaith and co-workers first investigated the UV aging effect on the stability of m-TiO 2 PSCs. [41] The PSCs were exposed to the simulated AM 1.5 sunlight. They found that when a UV filter was used, the PSCs were more stable. However, the encapsulated device was unstable when exposed to unfiltered simulated sunlight. They studied the surface chemistry of TiO 2 and proposed the mechanism responsible for photocurrent degradation in the solar cells. Upon UV light exposure, an electron-hole pair will be formed after the excitation of TiO 2 by UV light with energy higher than the bandgap of TiO 2 . Then photogenerated holes in m-TiO 2 react with oxygen absorbed at surface oxygen vacancies, which then become deep traps leading to charge recombination. As TiO 2 has a strong ability to extract electrons from organic materials as photocatalysts and from iodide anion (I À ) as electrodes in dye-sensitized solar cells, the driving force of decomposition was thought to result from the effect of electron extraction by TiO 2 from I À .
Ito et al. showed the degradation scheme of CH 3 NH 3 PbI 3 PSCs resulting from TiO 2 during light exposure in Figure 3a, and the relevant equations are as follows.
It speculated that TiO 2 can extract electrons from I À , resulting in the I 2 product at the surface of TiO 2 (Equation (2)), which deconstructs the perovskite crystal (Equation (3)). Then, the extracted electron that returned from the TiO 2 surface reacted with H þ , which resulted into further decomposition of the perovskite (Equation (4)). [42] Similarly, Wang and co-workers deemed that the emergence of I 2 under UV light soaking is the main reason affecting stability. [43] The related equations have shown that the electrons in the CB of TiO 2 could induce the oxygen adsorbed on the vacancies of TiO 2 into hydroxyl radicals and H 2 O under UV light soaking, leading to a loss of CH 3 NH 3 þ , as shown in Equation (7) and (8). Meanwhile, both hydroxyl radicals and holes could oxidize I À into I 2 , as shown in Equation (9) and (10).
Li and co-workers proposed a two-stage degradation process (Figure 3b) of TiO 2 -based PSCs under continuous UV irradiation in an inert atmosphere. [44] After excitation by UV light, the VB electrons of TiO 2 transit to the CB, leaving free holes (step 1).  [42] Copyright 2014, American Chemical Society. b) Li et al, Reproduced with permission. [44] Copyright 2020, Elsevier. c) Photographs of the mixtures of perovskite single-crystal powder and SnO 2 nanoparticle under UV illumination and d) the schematic diagram of the decomposition of PbX 2 under photocatalysis. Reproduced with permission. [45] Copyright 2019, Elsevier.

Photocatalytic Degradation of PSCs Induced by SnO 2
Although SnO 2 is considered as an ETL with inferior photocatalytic activity compared with TiO 2 , it is also a cause for serious destruction of the perovskite under UV illumination. Hang et al. investigated the UV photocatalysis degradation of perovskite induced by SnO 2 . [45] They uniformly mixed three types of perovskite single-crystal powders (MAPbI 3 , MAPbBr 3 , and MAPbCl 3 ) with SnO 2 nanoparticles and exposed them to continuous UV illumination for several days (Figure 3c). The degrees of perovskite degradation under UV can be observed from the color of the powder. The X-ray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS) measurements were carried out to precisely elucidate the effect of photocatalysis for the stability of MAPbI 3 , MAPbBr 3 , and MAPbCl 3 perovskite. It was observed that PbX 2 (X ¼ I, Br, Cl) and Pb 0 product in the MAPbI 3 :SnO 2 case increased much faster than those in MAPbBr 3 :SnO 2 and MAPbCl 3 :SnO 2 cases.
Based on the earlier results, they proposed that the degradation procedure of MAPbI 3 under SnO 2 photocatalysis is expressed as the following.
In the first stage of degradation (Equation (15)), the perovskite decomposed and PbI 2 product generated. Then, the PbI 2 can be further decomposed into Pb 0 and I 2 (Equation (16)) through interfacial redox reactions of electron-hole pairs, which are generated upon bandgap excitation when SnO 2 is exposed to UV light (Figure 3d). It is known that both Pb 0 and I 2 severely degrade the performance of PSCs as well as their long-term photostability. In addition, they demonstrated that the chlorine-based perovskite possesses the best tolerance for photocatalysis compared with the bromine-based and iodine-based perovskite owing to the strong Pb-Cl bond.

Strategies
The photocatalytic degradation of PSCs needs two basic factors: UV light and ETLs with photocatalytic activity. Thus, various types of strategies were proposed: 1) removing the UV spectrum through using the UV-filter and down-conversion luminescent material or simply using white LED lamps in laboratory studies; 2) strengthening the perovskite/ETL interface to resist the UV photocatalytic effect; 3) blocking the photocatalytic effect through inserting an interface with no photocatalytic activity; and 4) replacing the TiO 2 (SnO 2 ) by other ETLs with inferior or no photocatalytic activity.

Using White LED lamps
Various reports have shown that the stability of these devices was assessed under different light sources. These include xenon arc, [46,47] metal halide, [48] sulfur plasma, [49] tungsten halogen, and combinations of LED lamps. [25,50,51] Among them, only xenon arc and metal halide emit a significant amount of UV irradiation and their emission spectrum can be tuned to match the actual daylight spectral distribution. Based on our survey of the reported long-term stability profiles, most studies using TiO 2 or SnO 2 as ETLs used white LEDs to monitor device stability. It infers the importance of avoiding the photocatalytic activity of MO for stable operation of PSCs. The work by Burschka et al. was the first demonstration that MAPbI 3 -sensitized TiO 2 solar cells are stable up to 500 h (20% decay in PCE) in argon atmosphere and under maximum power point (MPP) tracking, using a white LED with 100 mW cm À2 . Up to now, using a white LED to assess the photostability of PSCs is the means in the laboratory studies.

Filtering UV Light
The commercial UV filter can effectively filter the UV light between 275 and 400 nm that cannot be absorbed by the Earth's atmosphere. Using a UV filter has been a general and simple strategy that avoids inducing the photocatalytic activity of TiO 2 and SnO 2 . To date, most PSCs can only work stably under illumination with UV filters. With the application of the UV filter, PSCs have achieved lifetimes of 10 000 h under simulated AM 1.5 G solar illumination at a stabilized temperature of 55 C (Figure 4a), which is equivalent to the total irradiation of 10 years of outdoor use in most of Europe. [52] Another approach to eliminate the UV light to the MO layer is directly coating a UV absorber layer on front of the transparent electrodes, such as ITO or fluorine-doped tin oxide (FTO). Ding and co-workers introduced an optimal UV absorber (2-(2H-benzotriazol-2-yl)-4,6-bis(1-methyl-1-phenylethyl)phenol, UV-234) on the front side of the FTO glass to filter the UV light between 300 and 400 nm (Figure 4b,c). [53] The device with a UV absorber layer exhibited an enhanced photostability as there was no PbI 2 peak at 12.71 after the UV aging test for 12 h ( Figure 4d). However, the device lost 2.2% PCE mainly because of a 2.7% decrease in J sc (21.43 vs. 20.85 mA cm À2 ) without the contribution of UV light to efficiency. TiO 2 nanoparticles/ graphene nanodot (TiO 2 NPs/GND) composites were used as a UV absorber layer placed in front of the PSC. [54] As shown in Figure 4e, GNDs with low oxygen concentration have resulted in the narrowed bandgap of TiO 2 through carbon doping or the presence of additional energy states, which lead to the enhanced UV absorption (Figure 4f ). As a result, the PSC with the TiO 2 NPs/GNDs showed only 15% decrease after 100 h of UV irradiation, whereas the PSCs with the TiO 2 absorber layer and without a UV absorber degraded 70% and 92%, respectively. The related stability tests are summarized in Table 1.

Inserting an Interfacial Modifier
The interfaces can block the photocatalytic activity of TiO 2 and SnO 2 to prevent the perovskite decomposition and device efficiency loss. In the following sections, we will summarize several  [52] Copyright 2017, the Author(s). Published by Springer Nature. b) Illustration of the device structure with UV absorber. c) EQE curves of the device with and without coating a UV absorber layer. d) XRD patterns of perovskite films with and without the UV absorber layer upon UV irradiation environments without encapsulation, Reproduced with permission. [53] Copyright 2017, Royal Society of Chemistry. e) Schematic illustrations of the charge-transfer process occurring between TiO 2 nanoparticles and GNDs, and f ) UV-vis absorption spectra of pristine TiO 2 NPs, GNDs, and TiO 2 nanoparticle (NP)/GND composites in water. Reproduced with permission. [54] Copyright 2018, American Chemical Society. types of interfacial materials which were inserted between these ETLs and the perovskite absorber. The reported interfacial modification materials and the related light stability tests, discussed in this review, are summarized in Table 2 as MgO, PbO, SrO, and so on) has been used to improve the device stability under UV irradiation. For example, Wang and co-workers introduced MgO or PbO as a surface modification layer into PSCs to block the contact of perovskite/TiO 2 and protect the perovskite from UV photocatalytic corrosion. [55,56] They demonstrated that PbO could serve as perovskite crystal seeds and induce a dense and robust perovskite film, which results in superior UV stability. [56] After recording the UV stability of devices under 364 mW cm À2 UV light, the PCE of PbO PSCs maintained 60% of initial value within 180 min, which was better than that of bare TiO 2 cells (Figure 5a). The SrO interlayer was also used to passivate the degradation sites at the perovskite/m-TiO 2 interface, [57] which is illustrated using spatially resolved μphotoluminescence (PL) (Figure 5b). The SrO-incorporated film exhibited almost no change in the PL peak intensity and peak position, which was attributed to the passivated degradation site. These devices with and without the SrO coating maintained 60% and 35%, respectively, of their initial efficiencies after 100 h of UV exposure. Moreover, Tavakoli et al. believed that a thin amorphous layer of SnO 2 (a-SnO 2 ) could shield the UV-induced charge transfer passes at the interface and improve the stability of the device under UV light. [58] Consequently, the double-layer structures of c-TiO 2 /a-SnO 2 ETL and m-TiO 2 /a-SnO 2 ETL have been introduced to modify the interface energetics, resulting in improved charge collection and decreased carrier recombination in PSCs. Among them, the UV stability of m-TiO 2 /a-SnO 2 -based perovskite device was improved drastically compared with pure TiO 2 , [59] whereas the PCE loss was only 3% after 60 h exposure to a UV lamp (Figure 5c).In addition, the inorganic perovskite also has been used to modify TiO 2 . The in situ modification protocol was proposed to form a compact CaTiO 3 layer on TiO 2 , which hinders the penetration of H 2 O/O 2 in the film. [60] The CaTiO 3 layer not only retards the recombination between electrons in TiO 2 and holes in the perovskite due to higher energy level of the CB bottom than TiO 2 , but also retards the photocatalysis activity of TiO 2 with less ".O 2 À " (Figure 5d). However, the high sintering temperature (450 C) of CaTiO 3 and lower conductive property than TiO 2 may limit their application.
Sulfide Materials: Ito and co-workers first introduced a blocking layer of Sb 2 S 3 between m-TiO 2 and CH 3 NH 3 PbI 3 to enhance the stability against light exposure. [42] Under UV irradiation, the holes generated in the TiO 2 CB can extract electrons from an I À anion in CH 3 NH 3 PbI 3 , resulting in perovskite crystal decomposition. With the Sb 2 S 3 layer, the charge extraction process can be prohibited, which makes CH 3 NH 3 þ fixed in the perovskite crystal and improves UV light stability but results in the enhancement of the charge recombination and a decreased V oc . The MAPbI 3 PSCs without Sb 2 S 3 interlayer deteriorated to zero efficiency within 12 h, whereas those with Sb 2 S 3 maintained 65% of initial efficiency (Figure 5e). Similarly, a uniform CdS shell was coated onto the surface of an m-TiO 2 layer. [61] CdS could suppress the intrinsic trap sites of TiO 2 originating from the oxygen vacancies, which create some energy levels lower than the ordinary CB level of TiO 2 and lead to charge losses by recombination. [61] As a result, the proposed TiO 2 /CdS PSCs achieved improved light stability, maintaining 77% of the initial efficiency after 12 h of fullsunlight illumination, while only 42% was retained for the TiO 2 PSCs. In terms of Cd-containing compounds, Cd 2þ cations and SO 4 2À anions have also been used to deposit a CdS layer on the TiO 2 surface by a chemical solution technique. [62] As a result, the device with passivated TiO 2 showed 20% performance loss, while the pristine TiO 2 device showed 60% performance loss after 60 h of illumination.
Zheng and co-workers converted ZnO surface into ZnS at the ZnO/perovskite interface through sulfidation (Figure 5f ). [63] A strong coordination between S and Pb 2þ creates a novel pathway for electron transport and reduces interfacial charge recombination. It slightly reduced the ETL transmission in the range of 310-380 nm (UV region), which benefited obtaining better photostability. ZnS-based PSCs displayed an outstanding improvement of UV stability by retaining 87% of its initial PCE over 500 h under UV irradiation, compared with ZnO-based PSCs of 0.07%. As one of the chalcogen compounds, ZnSe is a direct bandgap material of 2.8 eV, enabling broad absorbance in the UV region, which retards the UV degradation of perovskite materials without obviously decaying light harvesting in the rest of the solar spectrum (Figure 5g). The thin layer of ZnSe cooperated with TiO 2 can effectively reduce the interfacial charge recombination and promote carrier transfer because of the cascading CB level. [64] Finally, the ZnSe-based device exhibited superior photostability, which retained 90% of their initial efficiency after aging for 500 h.
Halide Materials: The halides are raw materials for preparing perovskite light absorption layers. [65] Researchers found that the halide could enhance the contact of perovskite with TiO 2 or SnO 2 and suppress the interfacial recombination. Wang and co-workers studied the impact of CsBr or CsCl interface modifier for c-TiO 2 . [43,66] As for CsBr modification, after 95 min UV aging of 523 mW cm À2 , the normalized absorbance for the perovskite film on glass without TiO 2 or on c-TiO 2 /CsBr substrate was at 0.99, while the absorbance of the perovskite film on c-TiO 2 fell sharply to 0.85 (Figure 6a). The better UV stability can be attributed to CsBr located at the interface covering UV-induced catalytic reaction activity sites of TiO 2 . Recently, the CsI-SnO 2 complex was used as an ETL for the stable PSCs by Xu et al. [67] However, they emphasized that the incorporation of Cs þ into the perovskite at the bottom interface enhances the resistance against UV illumination. As shown in Figure 6b, the PbI 2 peaks (28.2 ) of the pristine perovskite belong to UVinduced decomposition after 500 h of UV aging, while part of the perovskite film's color changes from black to yellow. The UV aging test results are shown in Table 2.
Fluorine (F) has the highest electronegativity among all elements. It has been reported that the incorporation of F anion  [56] Copyright 2016, Royal Society of Chemistry. b) μ-PL measurement results for pristine and 84 h UV-exposed perovskite films: left is PL intensity and right is PL peak position. Reproduced with permission. [57] Copyright 2018, Elsevier. c) UV stability of PSCs based on m-TiO 2 and m-TiO 2 /a-SnO 2 . Reproduced with permission. [59] Copyright 2018, American Chemical Society. d) Schematic for the retarded photocatalysis effect of TiO 2 after in situ surface modification. Reproduced with permission. [60] Copyright 2019, AIP Publishing. e) Photoenergy conversion efficiencies of solar cells during light exposure without (green) and with (black) Sb 2 S 3 . Reproduced with permission. [42] Copyright 2014, American Chemical Society. f ) Illustration of ZnO-ZnS-based PSC device. Reproduced with permission. [63] Copyright 2019, American Chemical Society. g) UV-vis spectra of the optimized TiO 2 and ZnSe film; inset shows the normalized PCE decay of devices based on ZnSe and TiO 2 as a function of storage time upon UV irradiation. Reproduced with permission. [64] Copyright 2018, American Chemical Society.
www.advancedsciencenews.com www.advenergysustres.com (F À ) in the perovskite layer led to passivation of both anion and cation vacancies of perovskite and consequently improved PSCs' device efficiency along with enhanced thermal and environmental stability. TiF 4 was doped into a TiCl 4 precursor solution by a one-step nonhydrolytic sol-gel method to obtain in situ lowtemperature solution-processed TiO 2 nanocrystals. [68] TiF 4 with stronger Ti-F ionic bonds is much more stable than TiCl 4 because of the much larger electronegativity of F than chlorine (Cl). The formation of Ti-F bonds brings increased electron mobility, decreased density of electronic trap states, and inhibits the photocatalytic activity of TiO 2 due to preferentially binding of F atom onto the (001) facet of anatase TiO 2 . As the photocatalytic activity of TiO 2 is sensitively dependent on the exposed crystalline facet and surface fluorine termination, the F doping-induced exposure of (001) facet is expected to weaken the photocatalytic activity of TiO 2 . Finally, the 12% F-TiO 2 device retained 68% of the initial PCE after 26 h of continuous UV irradiation, whereas PCE of the control device degraded to nearly zero after only 10 h. Hang et al. introduced a uniform NH 4 Cl film on SnO 2 . [45] A chlorine-rich mixed-halide perovskite MAPbI x Cl 3Àx interlayer (ClMPI) was formed through halide exchange between MAPbI 3 and NH 4 Cl (Figure 6c). It was demonstrated that ClMPI possessed excellent tolerance to photocatalysis owing to the strong Pb-Cl bond. After exposing the PSCs to 365 nm UV light with a high power of 100 mW cm À2 , the ClMPI PSCs could retain over 82% of the original value after 500 h, while the control device showed a constant drop and reduced to <10% of its original value within 100 h (Figure 6d). During their MPP test, the ClMPI-based device can maintain >80% after 500 h in a nitrogen glove box under continuous 1 sun illumination without any UV filter (Figure 6e), which was two orders of magnitude improvement compared with the control devices (only 4 h).
Very recently, Seok et al. introduced an FASnCl x interlayer between SnO 2 ETL and perovskite layer, through coupling Clbonded SnO 2 (Cl-bSO) with a Cl-containing perovskite precursor. [1] This interlayer can enhance charge extraction and transport from the perovskite layer, and fewer interfacial defects, which lead to a record PCE of 25.8%. The efficiency of PSCs based on a planar TiO 2 ETL (Cl-TiO 2 ) decreased substantially over time, even within an hour, in contrast to those based on Cl-bSO. The main reason is the different number of bound Cl À ions at the interface. The long-term operational stability of the unencapsulated device was tested under MPP tracking without a UV cutoff filter. As a result, the device maintained roughly 90% of its initial efficiency after 500 h. They emphasized that interfacial binding between the perovskite and electrode in the planar structure had substantial influence on the stability, rather than being a unique property of SnO 2 .

Inserting an Organic Interface
The fullerene and its derivatives have excellent semiconductor properties and they were widely used as interlayers between MO and the perovskite to cut off the photocatalytic degradation Figure 6. a) Normalized absorptance of perovskite, TiO 2 /perovskite, and TiO 2 /CsBr/perovskite after UV aging. Reproduced with permission. [66] Copyright 2016, Royal Society of Chemistry. b) XRD patterns and the corresponding photos of pristine and UV-aged perovskite films based on SnO 2 and CsI-SnO 2 , Reproduced with permission. [67] Copyright 2021, Wiley-VCH GmbH. c) Schematic illustration of the formation of the MAPbI x Cl 3Àx interlayer. d) Continuous MPP tracking for 200 h of PSCs in a nitrogen glove box under 1-sun illumination with a 420-nm cutoff UV filter and 500 h without any filter, and e) pure UV light aging test of control and ClMPI-based PSCs for 500 h. Reproduced with permission. [45] Copyright 2019, Cell Press.
Unfortunately, the introduction of the fullerene used as an interlayer to block the photocatalysis effect of SnO 2 usually leads to rapid initial degradation, which is known as a "burn-in" effect. Hang et al. found that introducing a small organic molecule bathophenanthroline (Bphen) can largely stabilize PCBM through forming thermodynamically stable complexes under UV and natural light soaking. [76] For the light-soaking test, these various types of devices were measured at 60 C for 1000 h of continuous MPP tracking without any UV filter (Figure 7b). After light soaking, the PCE of the PCBM-based device maintained stabilized efficiency of over 70% of its initial efficiency, and the PCBM: Bphen-based device retained over 92%, which was far beyond the SnO 2 device (only remaining <30% within 100 h). As shown in Figure 7c, the PCBM:Bphen device retained over 95% of its initial efficiency after 1100 h of the UV preconditioning test, which was far beyond the SnO 2 -based device (nearly zero) and exceeded the PCBM-based device (maintaining 70% of its initial PCE).
One of the natural selecting sunscreens of plants-sinapoyl malate (SM) as an ester derivative of sinapic acid has been used to modify the interface of TiO 2 /perovskite (Figure 7d). [77] The SM molecules formed a protonated model to accomplish the binding of SM 2À with TiO 2 (TiO 2 -SM), exhibiting a broader absorption spectrum below 400 nm and lower transmission in the range of 300-390 nm (Figure 7e), which decreases the destruction by UV radiation for PSCs. As a result, the PSC device with TiO 2 showed rapid decomposition with almost whole performance decay within 500 h under UV illumination. In comparison, as expected, the cells with SM treatment retained over 90% of its original efficiency after illumination. In addition, polyethyleneimine ethoxylated (PEIE) was proposed to improve the UV stability by blocking the transformation of Ti 3þ -V O states (as mentioned earlier).  [73] Copyright 2015, American Chemical Society. b) Light-soaking test of the SnO 2 , SnO 2 /PCBM, and SnO 2 /PCBM:Bphen devices and c) long-term 365 nm UV stability. Reproduced with permission. [76] Copyright 2021, John Wiley and Sons. d) Schematic illustration of SM assembled at the interface between TiO 2 and perovskite and e) transmission spectra of TiO 2 films modified by SM. Reproduced with permission. [77] Copyright 2018, John Wiley and Sons. The optimized devices maintained %75% of its initial PCE (20.51%) under UV irradiation at 72 days, whereas the normal devices failed completely.

Component Optimization of Perovskite
In case of perovskite degradation, it has been analyzed by considering the local electric field which is generally the practical working condition for solar cells and often causes ion migration as it can accelerate the reaction between oxygen and perovskite in light conditions. [78,79] Seok et al. introduced methylenediammonium dichloride (MDACl 2 ) for stabilizing the a-FAPbI 3 phase based on a typical m-TiO 2 structure. [80] The optimized device maintained %90% of its initial efficiency (>23.0%) of over 600 h of full-sunlight illumination without UV cutoff. They attributed the excellent photostability to the high concentration of Cl ions in the TiO 2 /perovskite interface and the stable black phase of a-FAPbI 3 .
Han et al. used the bifunctional organic molecular (5-ammoniumvaleric acid, 5-AVA) to inhibit the MAI loss at grain boundaries, crystal reconstruction, and irreversible ionic migration of MAPbI 3 under multiple actions of light, heat, and electrical bias. [21] It was illustrated that the ammonium groups and carboxyl groups of 5-AVAI make the perovskite grains connect with each other or connect the MO (Figure 8a). The printable PSCs filled with (5-AVA) X MA 1ÀX PbI 3 showed a slower attenuation process under the UV preconditioning test (Figure 8b) and they kept their performance after 1000 h of continuous light soaking at MPP (Figure 8c). Notably, a printable mesoscopic PSC passed IEC61215:2016 qualification tests with more than 9000 h of operational tracking, which indicates that the homologous crosslinking reaction between two selected molecules is a useful and meaningful approach to improve the PSCs' stability.
A "sunscreen" strategy using a tautomeric molecule is also proposed to regulate the defect passivation and enhance UV photostability. 2-Hydroxy-4-methoxybenzophenone, having the hydrogen on hydroxyl and the adjacent oxygen on carbonyl, can act as a sunscreen UV absorber. The molecule was demonstrated to form stable coordination and restrain UV degradation by tautomeric transition (Figure 8d), which induces high defect formation energy (À1.35 eV). [81] It was demonstrated that the carbonyl groups form an intermediate adduct with PbI 2 . The sunscreen-based PSC exhibited excellent UV stability at 285 and 365 nm ( Table 2). As a widely used UV absorber in the industry, 2-(2-hydroxy-5-methylphenyl)benzotriazole (UVP) was also used as an additive to PbI 2 . [82] The incorporated UVP molecule absorbs UV through the opening and closing of the chelating ring. The comparisons of stability are presented in Table 2. Reproduced with permission. [21] Copyright 2020, Elsevier. d) Schematic illustration of the sunscreen PSCs. Reproduced with permission. [81] Copyright 2021, John Wiley and Sons.

UV-Inactive Inorganic ETLs
Replacing TiO 2 by other semiconductor materials with inferior or even no photocatalytic effect can effectively improve the photostability. Snaith and co-workers showed replacement of the mesoporous n-type TiO 2 by an insulating mesoporous Al 2 O 3 scaffold. [41] The result showed that the stability of the mesoporous Al 2 O 3 device was much better than that of the TiO 2 device under UV light. The encapsulated Al 2 O 3 -based solar cells with epoxy resin and a glass coverslip in a nitrogen-filled glove box display stable photocurrents under continuous full-spectrum sunlight for a period of over 1000 h (Table 3). Then, they used Al-doped TiO 2 to reduce the nonstoichiometric oxygen-induced defects in TiO 2 . [83] This type of Al(III) substitutional doping combined with oxygen vacancies does not introduce defect energy levels in the bandgap because of the stable threefold coordination of Al(III). The stability enhancement is attributed to the substitutional incorporation of Al in the anatase lattice. Hematite (α-Fe 2 O 3 ) has the most stable iron oxide with n-type semiconducting properties under ambient conditions and compatible energy band structure with PSCs. [84] It has a low photocatalytic activity due to the high recombination rate of electrons and holes, as well as low diffusion lengths of holes, which could be beneficial for superior UV stability. In comparison with the TiO 2 , α-Fe 2 O 3 (%2.3 eV) with a narrow bandgap causes parasitic light absorption in the visible light region below 600 nm. [85,86] The use of a compact α-Fe 2 O 3 /discrete α-Fe 2 O 3 nanoisland structure builds some gaps, which ensure visible light transmittance utilized by the perovskite layer and enhanced UV light absorption. [87] Then, Ni-doped and N, S codoped graphene quantum dots (NSGQDs) were sequentially used for Fe 2 O 3 -based PSCs, [88] which improved the electron conductivity and tailored multiple interfaces, respectively, as well as UV light stability (Figure 9a). Nb 2 O 5 has high optical transparency, chemical stability, outstanding conductivity, as well as matched energy levels to perovskites, and it has been used in both dye-sensitized solar cells and PSCs by the sputtering technique. [89] Wang et al. reported a low-temperature solution route to prepare Nb 2 O 5 nanoparticles and further utilized them as ETLs (Figure 9b). [90] The Nb 2 O 5 nanoparticles show much better anti-UV ability due to better chemical stability, which suppressed the UV light-induced O 2À (as mentioned earlier) and showed improved UV-stable perovskites. The normalized J sc of unencapsulated PSCs was recorded upon UV exposure (Figure 9c). Devices based on Nb 2 O 5 can retain 93% of their initial J sc , whereas those devices based on TiO 2 can only retain 40% of their initial J sc . Finally, PSCs with improved UV stability were obtained by replacing TiO 2 with Nb 2 O 5 (Table 3). [90] Some perovskite-structured semiconductors have been used as an UV-inert ETL, but the high crystallization temperature (>1000 C) inhibits their application on glass substrates. Figure 9. a) Normalized PCE decay of perovskite devices based on different ETLs as a function of UV irradiation time (500 mW cm À2 ). Reproduced with permission. [88] Copyright 2020, American Chemical Society. b) Degradation diagrams of Nb 2 O 5 -and TiO 2 -based PSCs upon UV exposure and c) normalized J sc of Nb 2 O 5 and TiO 2 -based PSCs. Reproduced with permission. [90] Copyright 2019, American Chemical Society. d) ZTO particle, e) the photocatalytic degradation of MB on different catalysts under full-spectral light, and f ) photostability tests under UV illumination for PSCs based on ZTO and TiO 2 ETLs. Reproduced with permission. [92] Copyright 2019, John Wiley and Sons. g) PSC structure containing 2D Ti 1Àδ O 2 as an ETL, and h) the effects of UV irradiation on carrier dynamics of c-TiO 2 and 2D Ti 1Àδ O 2 : amplitude (ΔA rec ) correlates with the number of carriers relaxed through recombination. Reproduced with permission. [122] Copyright 2019, American Chemical Society.
www.advancedsciencenews.com www.advenergysustres.com La-doped BaSnO 3 (LBSO) has high electrical mobility (320 cm 2 V À1 s À1 ) at room temperature as well as inferior UV photocatalytic ability because of its small dipole moment, ascribed to the cubic perovskite structure without octahedra tilting. Seok et al. first obtained a compact, crystalline LBSO thin film below at least 500 C by a crystalline superoxide-molecular cluster colloidal solution. [91] The monitored PCEs of the encapsulated devices were tested with a metal-halide lamp, including UV radiation. The TiO 2 -based device completely degraded within 500 h, whereas the LBSO device maintained 93.3% of the initial PCE value after 1000 h (Table 3), which showed superior resistance against photodegradation. ZnTiO 3 (ZTO) is a perovskite-structured semiconductor material with a wide bandgap (3.25 eV), excellent chemical stability, and poor photocatalysis. Wei et al. introduced ZTO as a UV-inert ETL to improve light stability (Figure 9d). [92] The photocatalytic activity of ZTO and TiO 2 was evaluated by the photocatalytic degradation of methylene blue (MB) aqueous solution under fullspectrum light irradiation in Figure 9e. MB drastically degraded by TiO 2 (70% loss after 20 min), while the ZTO sample has low photocatalytic activity on MB molecules (15% loss after 50 min). In contrast to the TiO 2 -based device with fast degradation, the ZTO-based device maintained 90% of its initial value after 100 h. The PCE evolvement of the device under UV light (365 nm) is shown in Figure 9f, showing the UV-inert property of this interface.
Low-dimensional materials have been used as ETLs in PSCs owing to their superior mechanical and electrical properties. The novel 2D atomic sheets of the titania (Ti 1Àδ O 2 ) layer (ASTL) were prepared by solution-processed atomic layer-by-layer deposition technique (Figure 9g). [93] As the 2D titania atomic sheets possessed the unique properties of high UV transparency and negligible oxygen vacancy, the device exhibited excellent UV stability as compared with that consisting of a conventional c-TiO 2 ETL. As shown in Figure 9h, upon UV light exposure, the recombination number (ΔA rec ) substantially increased by nearly 100% for the compact TiO 2 ETL, whereas it remained nearly constant in devices with the 2D Ti 1Àδ O 2 ASTL. The major property that differentiates UV sensitivity indicated that less photogenerated electrons and holes recombined at the perovskite/Ti 1Àδ O 2 ASTL interface. Similarly, the 2D TiS layer also exhibits lower photocatalytic activity due to the higher UV transparency and less sulfur vacancies. [94] In addition, a series of 1D TiO 2 nanocolumn photonic structures has been used as ETLs and these vertically aligned nanocolumn arrays exhibit better optical absorption in the UV spectrum compared with c-TiO 2 , in particular in the wavelength range of 300-370 nm. [95]

Organic Semiconductor Materials
Some nonmetallic oxides are potential replacements for MO as they achieve long-term light stability in both rigid and flexible PSCs. For example, the carbon-based material family, including fullerene (C 60 ) and their derivatives (PCBM), graphene, graphene QDs, etc., [75,96,97] has inspired much research effort (Table 3).
Fullerenes and their derivatives are among the most widely used n-type materials in organic electronic devices as they have both a suitable energy-level alignment and superior electron mobility. [98] They can also reduce the trap density, passivate grain boundaries of perovskite films, as well as lead to higher UV light stability of PSCs in contrast to m-TiO 2 . To realize high-efficiency and UV-stable C 60 -based PSCs, the Li-TFSI-doped C 60 or C 60 / ultrathin-TiO x (C 60 /u-TiO x ) bilayer was sequentially designed as a compact ETL by Liu et al. [99] The modified C 60 maintained less interfacial charge accumulation and excellent charge extraction at the electrode interface and significantly enhanced UV stability, as illustrated by Mott-Schottky (MS) plots (Figure 10a-c). Before UV illumination, the different ETL-based devices had the same order of magnitude (10 12 ). After UV illumination, the slope of the C 60 -based device (À8.24 Â 10 11 ) and TiO 2 -based device (À3.28 Â 10 10 ) changed to be lower than the C 60 /u-TiO x device (À1.85 Â 10 12 ). The higher slope of the C 60 /u-TiO x -based devices indicates less interfacial charge accumulation and excellent charge extraction upon UV irradiation. Finally, the C 60 /u-TiO x bilayer, C 60 -based, and TiO 2 -based PSCs retained 83%, 61%, and 0% of their initial performance after 312 h of UV irradiation, respectively.
PCBM is one of the most usual types of organic ETLs, which is generally used to obtain hysteresis-free n-i-p PSCs. [100,101] A low doping ratio of graphene QDs (GQDs) has been used to increase the low conductivity of PCBM to 0.422 mS cm À1 because of the long electron lifetime and ultrafast electron extraction. It was noted that GQDs could fill the electron traps that originated from the dimerization of PCBM upon light exposure, which minimizes the negative impact on PCBM dimerization. [102] As shown in Table 3, the device with PCBM:GQDs maintained >80% of its original value under continuous full-spectrum sunlight over 300 h, which is about 1.5 times than the control device. As for amine-based fullerene, PCBDAN has also been investigated to modify PCBM by a simple self-organization method. After 240 h of UV light soaking (190 mW cm À2 ), the devices with PCBM:PCBDAN ETLs maintained 85% of its original efficiency, whereas the TiO 2 -based device only maintained %30%.
Subsequently, a low-temperature (<100 C) prepared ternary ETL: C 60 and self-organized PFN (poly [ (9,9-bis(3 0 -(N,Ndimethylamino)propyl)-2,7-fluorene)-alt-2,7-(9,9-dioctylfluorene)] to PCBM (PCBM:C 60 /PFN) are shown in Figure 10d. [103] The C 60 additive is incorporated to effectively improve the conductivity of the ETL and the PFN is used to decrease the energy barrier at the ITO/ETL interface and improve the stability under full-sunlight illumination. Notably, the device based on PCBM:C60:PFN in Figure 10e shows slower degradation than the TiO 2 -based device during 1200 h of full-spectrum illumination. As one of the copolymer ETLs, the amino-functionalized copolymer semiconductor PFN-2TNDI has a conjugated backbone consisting of fluorine, naphthalene diimide, and thiophene spacers. [104] When recording the UV-light aging process (365 nm) in an inert-gas atmosphere, the PFN-2TNDI-modified devices showed a much slower decay than that based on bare TiO 2 upon 3000 h (Figure 10f ).

Downshifting Materials
The downshifting (DS) materials can efficiently absorb the UV photons and re-emit visible light that can be harvested by the photoactive layer. [105][106][107][108] Theoretically, the DS materials can www.advancedsciencenews.com www.advenergysustres.com not only reduce the destructive effect of UV on PSCs but also improve the performance of the devices. [106,107] The related parameters, discussed in this review, are summarized in Table 4. In 2016, the photopolymer V570-doped (fluorescent organic dye) fluoropolymeric layer has been coated on the front side of the PSCs. [109] The DS coating converts the UV range into visible light, which increases the photocurrent by 6% and prevents UV light from interacting with the perovskite. During the device stability test, the all-front-coated cells devices kept in an Ar-filled dry glove box and irradiated 8 h per day with a UV optical fiber lost 2% of their initial PCE after 3 months; in comparison, the uncoated cells dropped nearly to zero. After that, developing such routes that effectively harvests the short-wavelength UV light and converts it into lower-energy photons of the absorption spectrum of the perovskite layer has been extensively studied. Rare-Earth ions (such as Ce and Eu) have a charge transfer absorption band in the UV region, and their emission spectra are mainly located in the visible region. [110][111][112] Thus, those are commonly used as the luminescence center to enhance PSC performance and stability. [105] Rare lanthanide elements come into the view of DS materials as lanthanide complexes, which possess broadband absorption in the UV region and can effectively transfer the absorbed UV energy to lanthanide ions. One of the most efficient downconverting ions is europium (Eu 3þ ), as it has high-emission intensity. [112] Ling et al. provided a modified sol-gel method to obtain downconverted TiO 2 :Eu 3þ nanocrystal. [113] In the mesoporous TiO 2 :Eu 3þ structure, Eu 3þ ion is the luminescence center and the UV light absorption of TiO 2 nanocrystals can induce luminescence of Eu 3þ . The effective use of UV light showed an improvement of 29.2% from 12.22% to 15.79% in PCE. The same group also matched Eu 3þ ion with a phenanthroline derivative ligand (4,7-diphenyl-1,10phenanthroline). [114] The formed Eu complex has been coated on the reverse of FTO (Figure 11a), displaying an enhancement of J sc and PCE due to wide UV absorption (300-380 nm) and reemitting in the visible range (Figure 11b). When exposed to UV light at 320 nm, the Eu complex layer emitted red light (inset of Figure 11b) and exhibited excellent downconversion fluorescence emission properties. Under the UV lightsoaking test, Eu complexes acted as UV block layers and improved the maintained PCE from %50% to %62% after 10 h of UV illumination.
Similarly, Eu 2þ has also been used to form a phosphor material and it can replace a fraction of Sr sites, forming SrAl 2 O 4 :Eu 2þ , Dy 3þ (SAED). Cong et al. used the pulsed laser deposition approach to build a mesoporous SAED scaffold layer between c-TiO 2 and the perovskite layer, which avoids chemical ligands involved in nanomaterial synthesis. [115] The calculated UV light harvesting of the SAED-based film increases by 82% compared with the bare film due to the transition from the 5d to 4f level of Eu 2þ , [20] and it exhibits an emission in the green light region (Figure 11c). After UV irradiation, the devices with and without SAED maintained 92% and 40% of the initial PCE, respectively (Figure 11d). It demonstrates that SAED modification inhibited the light-induced deep trap states and the UV light-induced device degradation.
Some QDs have DS property with high PL quantum efficiencies (PLQE). CsPbCl 3 :Mn-based QDs were synthesized and applied onto the front side of the PSCs as the DS layer. [116] The high Figure 10. a) C 60 /u-TiO x bilayer, b) C 60 , and c) TiO 2 -based PSCs were obtained before and after UV irradiation at a 1 kHz probe frequency. Reproduced with permission. [99] Copyright 2019, Royal Society of Chemistry. d) Schematic illustration of the vertical phase separation in the ternary PCBM:C 60 :PFN blend and the molecular structure of PFN. e) Photoaging stability test under full-spectral illumination. Reproduced with permission. [103] Copyright 2018, Royal Society of Chemistry. f ) Normalized PCE as a function of testing time under UV illumination for the TiO 2 -and PFN-2TNDI-based PSCs. Reproduced with permission. [104] Copyright 2017, Royal Society of Chemistry.
www.advancedsciencenews.com www.advenergysustres.com PLQE (60%) and large Stokes shift (>200 nm) of CsPbCl 3 :Mn QDs can convert UV light into strong yellow emission (450-750 nm), as shown in Figure 11e. As a result, the CsPbCl 3 :Mnbased devices showed increased J sc (3.77%) and the maintained PCE was also improved by 12% after soaking in UV irradiation for 100 h (Figure 11f ). Furthermore, a thin layer of core-shell CdSe/CdS QDs (with 85% photoluminescence quantum yield) was placed on the back side of a device, [117] which possessed a wide absorption region (300-500 nm) and could convert high-energy photons into low-energy photons (around 620 nm). Similarly, CdSe/ZnS QDs (PLQE of 85%) with green emission were proposed as ETLs for the first time. [118] However, it needs a vacuum technique for perovskite film deposition to prevent the solvent to dissolve the underneath QD layer.

Conclusion and Perspective
Currently, all the state-of-the-art PSCs are generally based on the n-i-p conventional structure using TiO 2 or SnO 2 as ETL. One of the biggest challenges for these types of PSCs is that MO-based ETL causes photocatalytic activity under natural light and UV illumination. Although the phenomenon of photocatalysis-induced degradation is widely mentioned in numerous reports, only few studies have investigated the detailed degradation process and mechanism. Most results related to the degradation process and product were based on speculation rather than direct detection. More fundamental researches should be undertaken to deeply understand the origin of the photocatalytic degradation between the MO and perovskite, which can be a guide to develop a new strategy to solve this problem. We illuminate the main perspectives as follows. 1) Although the fundamental processes of TiO 2 and SnO 2 photocatalysis in liquid solution such as H 2 O have been extensively studied, the photocatalysis process in the solid perovskite could be different. Therefore, more photochemical mechanisms and basic principles of photocatalysis based on the solid phase should be investigated, aiming to provide important information for restraining TiO 2 or SnO 2 photocatalysis under real PSC operation. 2) Only the separated electrons or holes in MO that migrate to the surface successfully have chances to drive the photocatalysis reactions. Therefore, it should be helpful to inhibit photocatalysis through accelerating the electron/hole recombination process in MO or reducing their diffusion to interface.
3) The influence of the actual operation environmental factors, including temperature, humidity, and electronic bias, on the photocatalysis process should be concerned and evaluated. These factors will largely affect the photocatalytic rate and thus PSCs stability. 4) Fundamentally, the stability of perovskite absorption layer under the photocatalysis of TiO 2 or SnO 2 is dependent on their component Figure 11. a) Schematic illustration of enhanced photoelectric performance of PSC by the transparent Eu complex layer. b) PL excitation spectra of Eu complex layer with 0.5, 1.0, and 1.5 wt% content (inset shows the digital photograph of 1.5 wt% Eu complex layer when exposed to 320 nm UV light). Reproduced with permission. [114] Copyright 2017, American Chemical Society. c) The incident photo-to-current conversion efficiency (IPCE) spectra of PSCs with and without SAED (the inset picture shows the excitation and emission spectra of the SAED film). d) UV stability for control, inside, and outside of the SAED film-based PSCs under 365 nm UV light irradiation. Reproduced with permission. [115] Copyright 2017, John Wiley and Sons. e) Photographs of the fabricated CsPbCl 3 :xMnQD solutions and films with different Mn 2þ concentrations under 365 nm UV light, and f ) the aging test on the reference and CsPbCl 3 :0.1Mn QD layer under continuous UV irradiation (5 mW cm À2 ) in N 2 atmosphere for 100 h. Reproduced with permission. [116] Copyright 2017, American Chemical Society.
www.advancedsciencenews.com www.advenergysustres.com and structure. Especially, uncovering the differences on the photocatalytic degradation among Cs, FA, and MA perovskites is significant to guide the design of perovskite compositions. Photocatalytic degradation in MO-based cells could be simply inhibited by application of UV filters, but that would increase the additional cost of device production and seriously reduce the effective output by 10-20% due to the energy loss of the UV region when the device is in operation. It is known that solar cell efficiency is a key lever for PV cost reduction: a higher cell efficiency directly translates into a less-expensive PV system, reducing the levelized cost of electricity. Thus, filtering the UV will obviously reduce the cost performance of PSCs, which is generally regarded as a low-cost solar cell. Using inferior UV photocatalysis materials as ETLs can delay photocatalytic degradation; however, it is a still huge challenge to achieve efficiency as high as that of TiO 2 or SnO 2 PSCs. As for DS materials, light stability can be improved by blocking UV photons and re-emitting to visible light. However, the applicability on high-efficiency PSCs should be further verified as the downconversion efficiency of these materials is low. Recently, it has been encouraged that, through interface engineering, important breakthroughs have been achieved for improving the UV stability of PSCs in the high-efficiency device. Especially, introducing the high concentration of Cl ions at the interface between SnO 2 (or TiO 2 ) and perovskite can obviously resist the photocatalytic effect under UV light. These efficient devices exhibit excellent photostability after testing with MPP tracking under full solar illumination without a UV filter. Thus, we believe that interface engineering is the most promising manner to overcome the UV photocatalytic degradation of PSCs without compromising the efficiency. In addition, inserting the mesoporous insulator layer between ETLs (SnO 2 or TiO 2 ) and perovskite, such as ZrO 2 and Al 2 O 3 , is demonstrated to be an effective approach to improve the operational stability under continuous full-spectrum sunlight. [21,41] These device stabilities are promised to be further improved through replacing the metal electrodes by chemically stable carbon.
As the photocatalysis-induced degradation is triggered by the MO semiconductor transport layers in n-i-p PSCs, varying the device architecture will undoubtedly be an effective solution. Using polymer semiconductors without photocatalytic activity, such as PTAA, poly(4-butylphenyl-diphenylamine) (polyTPD), etc., as HTLs based on inverted (p-i-n) devices has achieved device operating stability. For example, Snaith and co-workers reported excellently long-term stable-inverted PSCs with PCE loss of less than 5% under full-spectrum-simulated sunlight in ambient atmosphere for more than 1200-1800 h at 70-85 C. [119,120] Huang and co-workers achieved p-i-n PSCs that maintained 97% of the initial efficiency after operation at the MPP under a plasma lamp with light intensity equivalent to AM 1.5 G without a UV filter for 1200 h at 65 C. [121] Thus, the UV stability of PSCs strongly depends on the device architecture.
Finally, despite the great emphasis laid on stability-related investigations, studies lack consistency in experimental procedures and parameters, especially the spectrum of light source. The UV photocatalytic degradation level of PSCs is seriously affected by light spectrum distribution, whose impact on device stability is still not fully understood or appreciated. It is difficult to compare the available results, mostly because of differences in the light source and measuring parameters. It is therefore challenging to reproduce and compare results and thereby develop a deep understanding of UV photocatalytic degradation mechanisms and the identification of various degradation factors.