Impact of H2O on organic-inorganic hybrid perovskite solar cells

The performance and stability of organic–inorganic hybrid perovskite solar cells (PSCs) is sensitive to water and moisture in an ambient environment. Understanding how H 2 O influences the perovskite material is also important for developing appropriate control strategies to mitigate the problem. Here we provide a comprehensive review on the effect of water on the state-of-the-art lead-based perovskite solar cells in terms of perovskite material design, perovskite film preparation, device fabrication, and photovoltaic application. It is found that a moderate amount of water can facilitate nucleation and crystallization of the perovskite material, resulting in better perovskite film quality and enhanced PSC performance. The perovskite materials are irreversibly destroyed by H 2 O after a certain level of water, but they exihibit better tolerance than initially expected. Humidity resistant fabrication of high-performance PSC devices and modules should therefore be favoured. Generally, water shows a negative effect on the long-term stability and lifetime of PSCs. To reduce the effects from water during outdoor operation, attention should be paid to different protection methods such as varying the perovskite composition, optimizing the electron/hole transport layer and encapsulation of the device.


Introduction
In 2009 an organic-inorganic hybrid perovskite material was incorporated into a dye-sensitized solar cell (DSSC) as a light absorber for the first time. Though the initial efficiency was low (3.8%), the new material exhibited good crystallinity, suitable optical properties, and unique electronic properties. 1 By optimizing the fabrication process, choosing new core materials, and designing a better photovoltaic device structure, the energy conversion efficiency of perovskite solar cells (PSCs) has recently achieved a value of 422%. 2 The light-harvesting perovskite material RAMX 3 commonly refers to a 3D organic-inorganic hybrid compound in which RA is a monovalent organic cation (methylammonium (CH 3 NH 3 + , MA + ), or formamidinium (HN = CHNH 3 + , FA + )), M is a bivalent metal (Pb 2+ , Sn 2+ ), and X is mostly a halide anion (Cl À , Br À or I À ). Capable of being mixed with cations (e.g. adding FA + into MAPbI 3 or partially replacing Pb 2+ with Sn 2+ ) or anions (e.g. mixing various halides) in crystal units, RAMX 3 can be adjusted to obtain a moderate bandgap (E g ) for a single junction solar cell (1.1-1.4 eV) or tandem solar cell (B1.8 eV). 3 For example, MAPbI 3 possesses the lowest E g of 1.6 eV. 4 A suitable bandgap could cover most of the visible light region from B400 nm to 800 nm, and potentially expand to the infrared region. 5 In addition, the diffusion length, carrier lifetime, and mobility, though these vary by composition, are apt for separation and collection of charge carriers in solar cell devices. RAMX 3 perovskites possess intrinsically simple growth kinetics, and PSCs resemble other thin film solar cells in terms of their device structures and materials to some extent, which make PSCs easy to be fabricated and potentially a low-cost photovoltaic technology. 6,7 The device architecture of PSCs can be divided into two distinct types: regular n-i-p PSCs and inverted p-i-n PSCs. In this review, unlike in most of the previous studies, we consider an integrated architecture shown in Fig. 1: electrode/second functional layer (SFL)/perovskite/ first functional layer (FFL)/transparent conductive oxide (TCO) substrate. Here, FFL refers to the functional layer between the perovskite and TCO substrate, and SFL refers to the functional layer between the perovskite and electrode. In the n-i-p architecture, the FFL can accept and transport photoexcited electrons, generally using metal oxides such as ZnO, TiO 2 (mesoporous and planar) etc. Correspondingly, the SFL acts as a hole selective acceptor. Inversely, in the p-i-n architecture the FFL is used to separate the excited holes and deliver them toward the TCO and the SFL is responsible for the transport of electrons.
Though the efficiency of PSCs is comparable to that of commercial solar cells (e.g. mc-Si, c-Si, CdTe), 8,9 one of the major barriers in front of their commercialization is the poor material and device stability. [10][11][12] A great deal of studies show that the PSC is very likely to degrade in an ambient environment. Moisture (or water) is one of the major degradation triggers for the perovskite material itself and the PSC.
In earlier studies, H 2 O was widely regarded as harmful to MAPbI 3 perovskite by turning it back to PbI 2 and MAI irreversibly, which results in loss of optical absorption and severe electronhole recombination. However, Zhou et al. in 2014 reported that annealing under humid conditions could actually greatly improve the film quality and its electronic behavior, demonstrating a positive role of water. 13 Later, many studies have proposed that H 2 O may have both advantageous and disadvantageous effects

Peter D. Lund
Dr Peter D. Lund is a Professor in Advanced Energy Systems at Aalto University, Finland. He has close to 40 years of experience in energy technologies, including solar and fuel cells. He has had visiting positions in China and Germany. Dr Lund is active in senior roles within European Union energy initiatives: he has chaired the Advisory Group Energy of European Commission and chairs the Energy Steering Panel of on RAMX 3 perovskites and PSCs. However, comprehensive understanding of the role of H 2 O in organic-inorganic hybrid perovskites and PSCs has seldom been discussed in previous reviews. Here we review the effect of water on the state-of-theart lead-based perovskite solar cells from a broad perspective, covering perovskite material design, perovskite film preparation, device fabrication, and application to pave the way for the development of high-performance and reliable PSCs. The structure of the review is illustrated in Fig. 2.
This review discusses the impact of water on lead-based perovskite solar cells in perovskite material preparation and operation stages, but also relevant strategies for improving their resistance against water. The preparation routes for the formation of perovskite films are considered with a discussion on how water in various states affects crystallization and film quality. Water-induced degradation routes will be comprehensively reviewed along with the final chemical products, and optical and electronic properties. Furthermore, based on the waterinduced degradation mechanisms, several approaches will be provided for better moisture resistance.

H 2 O in perovskite film preparation
Preparation of a perovskite film requires multiple steps: formation and collection of perovskite ingredients, preparation of perovskite precursors, and perovskite film growth, all of which provide opportunities for intrusion of H 2 O. Therefore, for a better understanding of possible routes for water penetration, we first describe the representative deposition methods with the related crystallization mechanisms.

Film preparation methods
A smooth, pinhole-free and highly crystalline RAMX 3 perovskite film is generally prepared by four representative processes (see Fig. 3A using MAPbI 3 as an example): one-step solution deposition (OSSD), sequential solution deposition (SSD), dual-source vapour deposition (DSVD), and vapour-assisted solution deposition (VASD). 1,7,[14][15][16][17] 2.1.1 One-step solution deposition (OSSD). OSSD is implemented with the first step that RAX (RA = methylammonium, formamidinium) and PbX 2 (X = I, Br, Cl) simultaneously at the mole ratio of 1 : 1 or 3 : 1 are added to polar solvents (DMF, DMSO, GBL, etc.). 1,18 After stirring constantly at a certain temperature, the clear solution is spin coated or drop coated on the first functional layer. This is followed by annealing the as-formed perovskite film to evaporate solvents and ensure the crystallization of RAPbX 3 .
Perovskite films in OSSD experience different crystallization growth mechanisms depending on the annealing temperature and annealing time. 19,20 When the substrate temperature is at or below 100 1C, the perovskite film experiences a multistage formation mechanism with three stages: the initial solution stage, the transition-to-solid film stage, and the transformation stage from intermediates into a crystalline perovskite film; however, when the substrate temperature is increased from 100 to 180 1C, the formation mechanism of the perovskite film is changed to the ''direct formation mechanism''. 19 An approximate time scale for each one of the stages and their evolution was established during the annealing of MAPbI 3 films at 100 1C in air, including the nucleation of small crystallites (i), a transitional stage during which a large number of grain boundaries are formed as precursors vanish (ii), an actual crystal growth period where different crystals coalesce (iii), and a final stage (iv) where the material is eventually formed. 20 In addition, a hot-casting technique at 180 1C was developed, during which the high substrate temperature at spin coating induces the formation of perovskite crystals. It was recently proposed that a Volmer-Weber growth mechanism occurs with island shaped grains, followed by integration into perovskite films. 21 2.1.2 Sequential solution deposition (SSD). SSD is a two-step deposition method in which RAX and PbX 2 are dissolved in solvents separately. 14 The PbI 2 DMF solution is spin-coated and forms a PbI 2 film at first, then transforms into a MAPbI 3 film by reacting with the MAI iso-propanol solution. Annealing is normally needed to completely transform the unreacted PbI 2 to MAPbI 3 .
Fu et al. reported two crystallization formation routes for SSD, depending on the concentration of MAI in iso-propanol (Fig. 4). 22 When the concentration of MAI is lower than 8 mg mL À1 , a solidliquid interfacial mechanism takes place and the as-formed MAPbI 3 film will gradually block the MA + from further reacting with the inner PbI 2 ; whereas, a dissolution-recrystallization pathway occurs at 410 mg mL À1 MAI solution, during which a quick formed MAPbI 3 film covers the PbI 2 surface immediately, demanding longer reaction time or higher MAI concentration to fulfill the reaction via the formation of Pb 4 2À . It should be also noted that similar competing pathways are proposed to control the growth of freestanding MAPbI 3 crystals, in situ transformation and dissolution-crystallization mechanisms. 23 2.1.3 Dual-source vapour deposition (DSVD). Vapor-based deposition techniques in which the ingredients are converted to gaseous states, reproducibly produce high-quality and dense perovskite films. DSVD, a dual chemical vapor source deposition technique, was developed by Snaith and his colleagues in which MAI and PbCl 2 precursor salts are co-evaporated under vacuum to deposit a MAPbI 3Àx Cl x film followed by further annealing to realize full crystallization. 15 Afterwards, a sequential evaporation    2.1.4 Vapour-assisted solution deposition (VASD). By combining the vapor deposition method and solution process, VASD conceptually inherits both advantages. A PbI 2 film is initially fabricated on the FFL by a solution process, and then it is reacted with the MAI vapor, leading to splendid crystallization and full coverage. 16 Recently, VASD has been further developed to operate under low pressure, showing shorter reaction time and lower MAI sublimation temperature. 28,29 In VASD, a onedimensional, top-down reaction route was proposed for the diffusion-controlled, strongly temperature-dependent crystal growth: 30 (1) MAI initially reacts with the surface of the PbI 2 film, leading to the formation of perovskite crystal nuclei; (2) MAI gas molecules further diffuse into the inner PbI 2 region, inducing the growth of more perovskite crystals perpendicular to the substrate; (3) after the reaction, the perovskite grains merge together, and appear as larger grains.

Role of H 2 O in perovskite film preparation
In the above methods, H 2 O in various forms can affect the perovskite reaction process and the crystallization dynamics, leading to desirable or undesirable changes in the perovskite crystals as well as grains and even the morphology and bulk phase of the perovskite films, as shown in Fig. 3B. The mechanisms via which water in the preparation processes affects perovskite films will be discussed in the following sections.

H 2 O in precursors.
Most alkylammonium salts as well as some lead salts PbX 2 are hydroscopic, leading to the possibility that hydration water is added into the perovskite precursors and influences the formation of perovskite if the material is not stored in an inert atmosphere or is not dried very well before use. 31,32 Most recently, the hydration water in 3MAI:1Pb(Ac) 2 ÁxH 2 O (x = 1.5) in OSSD was proposed to combine with the new product MAPbI 3 in the DMF precursor, forming MAPbI 3 ÁH 2 O in the reaction (eqn (1) and (2)), benefiting the morphology and electronic properties of perovskite films by reducing surface roughness and the nonradiative pathways. 33 Following this work, further study demonstrated the optimized ratio 3MAI : 1Pb(Ac) 2 Á1.5H 2 O, ascribed to the trade-off of reducing the non-radiative pathway and increasing porosity. 34 Similarly, the water of hydrate MAI, although it does not impact coverage, also lead to improved performance of PSCs by significantly enhancing the lifetime of excited carriers in the MAPbI 3 layer. 31 (1) Moreover, a similar way of influencing the perovskite precursors is to directly add a small amount of liquid water into the solvents. The liquid water, unlike most uncontrolled hydration water, is accurately calculated and added into the perovskite precursors. For instance, 0-10% volume ratio of deionized water was added into the MAPbI 3Àx Cl x perovskite precursors and the optimized 2 vol% water additive resulted in large grains with less voids in the morphology of the perovskite films, leading to a decrease in nonradiative trap states. 35 In this case, one possible influential mechanism is the formation of stable hydrates MAPbI 3Àx Cl x ÁnH 2 O and a lower stabilization energy in the upcoming annealing step. Meanwhile, the changes of the physical and chemical thermodynamic properties of mixed solvents like boiling point, solubility, and vapor pressure, affect the orientation growth of the perovskite crystals. 35 Almost at the same time, another study also reported that the same volume of water was added into the perovskite precursor, but the changes in the films were minor and the device performances were influenced slightly (Fig. 5). 36 However, the different results in these two almost identical MAPbI 3Àx Cl x experiments have not been further explained yet and we in passing point out that the underlying cause is likely the difference in the substrates since water molecules may disperse evenly on the more hydrophilic surface of PEDOT:PSS 35 37,38 In parallel to the reports of water in OSSD precursors, one of the pioneering studies in two-step deposition was performed by Fu et al. who reported that 3PbAc 2 ÁPbOÁH 2 O existed in the PbI 2 film when using Pb(Ac) 2 Á3H 2 O as an ingredient, but the pure tetragonal phase MAPbI 3 nanostructure growth is unaffected based on the dissolution-recrystallization growth mechanism shown in Fig. 4B. 22 In addition, an appropriate amount of liquid water was shown to have a positive effect on the growth of perovskite by affecting the crystallinity of the PbI 2 film or participating in the growth of MAPbX 3 . 39 A small amount of water has been used as an additive in PbI 2 DMF solution and MAI isopropanol solution respectively. 39,40 Both methods lead to larger grain size and better crystallinity, but the detailed mechanisms are not the same. Water in the PbI 2 solution modifies the properties of the H 2 O/DMF mixed solvent, the PbI 2 /PEDOT:PSS interface energy and the PbI 2 crystallinity, leading to a smooth, even PbI 2 film with the preferred (001) crystal plane parallel to the substrate. 39 Since perovskite crystals maintain the original orientation at the initial stage, 41 better PbI 2 crystallinity will be conducive to the perovskite film. In contrast, the water in the MAI solution is likely to assist the solid PbI 2 in reacting with the MAI, promote the crystallization of MAPbI 3 on the (110) plane and enlarge the grain size. 40 The increasing grain size might be explained by ingression of H 2 O in the formation process, through hydrogen bonding interaction with MAPbI 3 . It should be noted that a DMF solution could absorb water vapor. Clegg and his coworkers added controlled volumes of water into PbI 2 / DMF solutions to simulate the role of ambient moisture, and found that an increasing concentration of water not only reduced the overall device performance but also exaggerated the scan-rate and directional-dependent hysteresis and introduce new transient behaviors. However, addition of water also improved the longterm device stability. 42 But this time the water amount added was two orders of magnitude lower than in the other reports, 39,40 and the device performance did not in this case fit the results of Wu et al.'s work. 39 Full understanding of the differences in the performance trends reported still needs further work, but could probably be linked to the discrepancies in the fabrication processes in a glove box and ambient environments.
2.2.2 H 2 O in the spin-coating step. As mentioned above, the perovskite MAPbI 3 formation in OSSD involves nucleation and crystallization processes. According to the thin film growth theory and empirical data, a fast and high-density nucleation is expected so as to obtain better crystallization and surface morphology of the MAPbI 3Àx Cl x film, 43 as supported by some studies using physical or chemical means to increase the number of nucleation sites. [44][45][46] Therefore, reducing the humidity directly leads to high supersaturation, and thus sidesteps island growth so as to heighten the coverage of the film. 43 Nevertheless, unlike OSSD, moisture is more likely to participate in the reaction of PbI 2 film with MA + ions in the twostep spin-casting. Based on the interfacial reaction mechanism (Fig. 4A), compact MAPbI 3 covering PbI 2 hinders the 8 mg mL À1 MAI solution reacting with the inner PbI 2 . 22 One solution is to prewet the substrate before spinning PbI 2 by controlling the exposure of the films towards moisture. 47 The 3 min exposure of mesoporous TiO 2 films to moisture resulted in more conversion of PbI 2 and in turn improved device efficiency, by virtue of the porosity of the PbI 2 film into which MAI could infiltrate more easily. We propose that water molecules may disperse discretely on the TiO 2 surface because of its poor hydrophilicity and then these in-site H 2 O molecules attract PbI 2 by hydrogen bonds resulting in PbI 2 isolation, and thereby an increase in interspace is expected.
Also, moisture in the spin-casting process induces better reactions, larger grain size and better interconnectivity between perovskite crystals. Two mechanisms are responsible for the better perovskite crystallinity. For one, perovskite hydrates (MAPbI 3 ÁH 2 O, MA 4 PbI 6 Á2H 2 O and MAPbI 2 ClÁyH 2 O) appear as intermediates in or after spin-coating, which activate the reaction between PbI 2 and MAI, likewise with their intrinsic larger structural interspace providing extra fast paths for MAI diffusion. [48][49][50] The formation of MA 4 PbI 6 Á2H 2 O may be due to the coexistence of MAI and MAPbI 3 in the presence of excess water. 51 For another, moisture liquefies and ionizes the MAI, which exhibits more powerful capability of reacting with PbI 2 . 48,49 Nonetheless, increasing the relative humidity could greatly roughen the film surfaces, so it would to some extent cancel out the benefits of better crystallinity. 52 It is also noteworthy that methylammonium salts (MABr) liquefied in the moist air could not react with the PbI 2 and thus more MABr is required to ensure the complete chemical transformation of PbI 2 (see Fig. 6). 53 Balance of the effects is pertinent to the optimized range of R.H. but it appears to be very difficult since determination of the optimized R.H. is entangled with other parameters. Despite these issues, several groups have, respectively, used moisture (R.H. 30-60%) to drive the reaction of PbI 2 with MAI in the spin-casting process without a further thermal annealing step. 48,50,54 A dissolution and recrystallization mechanism at the perovskite grain edges is proposed. 54 Here, H 2 O molecules are absorbed on the surface void states and grain edges, then destroy the bonding between MA + and the Pb-I cage. The released MA + then reacts with the Pb-I cage again after water evaporates, achieving void-free, larger crystalline films. Similar improvement is obtained by other solvent vapors, e.g. chlorobenzene for ambient engineering. 55 These successes offer a new platform, like using a room-temperature synthesis process to further reduce the costs of device fabrication and allow more acceptance.
2.2.3 H 2 O in the annealing step. The thermal annealing process in solution deposition is an important step for the growth of MAPbX 3 crystals and the formation of polycrystalline perovskite films, accompanied by the evaporation of residual solvent. Meanwhile, by thermal annealing, the concern that water may get trapped in the structure units and affect the as-prepared perovskite films, is suggested to be dispelled. 56 In fact, a synergistic effect induced by a certain humidity and thermal annealing actually improves the perovskite film quality and device properties. In 2014, Zhou and colleagues unprecedentedly annealed MAPbI 3Àx Cl x films in a damp atmosphere (R.H. 30%), and consequently obtained larger crystal size and less nonradiative recombination. 13 A certain degree of moisture slows the crystal growth kinetic process by its effect on the solubility of the components and supersaturation of the perovskite, leading to a large grain size. 43,[57][58][59] Also, a large crystalline grain structure and domain size in a moist environment is proposed to block the further ingression of water vapor. 60 MAPbX 3 (X = I À , Br À , Cl À ) crystals likewise coalesce to a larger size by the effect of water that modifies the grain boundaries, diminishing the number of defects. 57,61,62 This may be due to surface effects as well as lattice distortion that trigger the grain edges to be sensitive to moisture.
The presence of small quantities of newly formed PbI 2 were later shown to benefit the electronic properties of grain boundaries in humid annealing, increasing the local conductivity through passivating traps at the grain boundaries and performing a selfdoping role. 63 In situ techniques detected amorphous materials containing hydroxyl ions and the increase of grain size as direct evidence of the effect of water on FAPbI 3 films. 56,64 Besides H 2 O, the annealing process under different solvent vapors such as DMF and DMSO, 65 has resulted in better perovskite crystallinity. However, the mechanisms may be different. DMF and DMSO can dissolve PbI 2 and MAI as well as the MAPbI 3 film. The model by Liu et al. 61 proposed that the formation of a liquid or quasi-liquid phase on the surface and void area of the perovskite film by liquid phase sintering, may be responsible for film densification in the annealing process. An appropriate amount of H 2 O in the organic solvent vapor is expected to enhance the surface solvation/dissolution of perovskite primary crystals, facilitating the integration and merging of initial perovskite crystals into larger grains and simultaneously healing the pinholes upon annealing. 66 In a humidified air annealing process, highly hygroscopic MA cations pull moisture from the environment. Water then partially dissolves the perovskite material and enlarges the MAPbI 3 crystals. 58 For a perovskite film spin-coated under low humidity and annealed under high humidity, a high nucleation density induced by high supersaturation appears in the spin-coating stage and causes layer growth of perovskite films, and the modest crystal growth under high humid conditions in the annealing stage can benefit the formation of perovskite films with better crystallinity and lower crystal defect density. 43 Meanwhile, MA vapor-involved annealing could also heal the perovskite film. 67 In this case well connected grains and less impurities were found, 67 which could be attributed to the potential formation of the MAPbI 3 ÁxCH 3 NH 2 intermediate phase. 68,69 However, the high humidity processing results in the appearance of more PbI 2 as insulating regions localized at the grain boundaries and within perovskite grains. 63 To restrain the likely adverse effects, the range of optimized R.H. had better be precisely controlled by considering other parameters like temperature scales, and ingredients. Regardless of the extremely humid conditions, annealing in a moderate humid atmosphere could be a self-improvement method for the perovskite device. Also, as many studies are continuously revealing the benefits that moisture induces in the annealing process in OSSD, it exhibits the capability of device fabrication under less restricted conditions.
The striking features like larger crystal size and less nonradiative recombination are still realized by post-treating perovskite films. Both the phenomena and essence are partially in common with the thermal annealing in a moist atmosphere. Defect density, mainly surface defects, decreases in accordance with less nonradiative decay after recrystallization of boundaries with or without thermal annealing. 70 Meanwhile, it is likely that MAI structurally becomes more mobile for better reacting with the remaining PbI 2 and excessive MAI could be removed as the result of being solubilized by water, both of which allow defects to be dispelled or filled in. 31 Very recently, Zhou et al. found deactivation of the perovskite surface under the effect of hydrogen bonds with uncoordinated iodide ions, like the deactivation by PCBM, 71 shifting the deep-level defects to shallow-level ones. Nevertheless, different from the long-term changes in other research studies, deactivation disappears after the escape of water vapor.
Concerning the incomplete reaction of MAI and PbI 2 in the fabrication, residual PbI 2 or MAI could be often expected. Recent studies revealed that the spatial distribution of remnant PbI 2 could affect the role of H 2 O on MAPbI 3 . Unreacted PbI 2 , close to the substrate TiO 2 , was reported to accelerate perovskite degradation during the exposure towards moisture 72 or the synergistic effect of moisture and illumination. 73 However, Petrus et al. found that the appearance rate of PbI 2 and monohydrate MAPbI 3 ÁH 2 O was slowed down with short-term exposure (R.H. 90%, o3 h). 74 They explained it with the passivation effect of PbI 2 at MAPbI 3 grain boundaries or terminations as Lei et al. did. 75 Interestingly, long-term exposure (R.H. 75%, 12 h) caused the same degradation degree between stoichiometric based devices and PbI 2 -excess devices. 74 Although a full explanation is still lacking, it is possibly ascribed to substrates that affect the degradation degree, 73 which we will discuss in more detail in Section 3. In contrast, a MAI-enrich perovskite film, though it showed poor pristine performance, demonstrated improved electronic properties and better crystallinity after recrystallization in a humid atmosphere and a more stable energy output. 74 2.2.4 H 2 O in the whole solution deposition process. Very few reports study the influence brought by the changes of water in the whole solution deposition process. Lv et al. investigated the effect of H 2 O and the assistance of solvents on the morphology of MAPbI 3 perovskite films and related device performances. By using DMAC as a solvent, they found that the most moderate humidity level is 28%. 76 2.2.5 H 2 O in or after vapor deposition. One of the pioneering works in VASD was done by Raga et al. who reported that H 2 O could facilitate the reaction of CH 3 NH 2 gas with the PbI 2 film. 77 With the aid of HI gas exposure, an ultra-smooth film was successfully formed under ambient air within several seconds and no high heating temperature was needed, as shown in Fig. 7.
Patel et al. exhibit the effect of ambient air (R.H. 40 AE 10% at 21 AE 1 1C) after the two-step thermal evaporation process. 78 In analogy to the effect of moisture as mentioned in OSSD and SSD, water vapor exposure makes unreacted MAI mobile, thereby facilitating chemical transformation of PbI 2 to MAPbI 3 . Intermediate dihydrate MA 4 PbI 6 Á2H 2 O was also detected, and it consumed excess MAI and favored MAPbI 3 growth (Fig. 8).

Effect of H 2 O on perovskite composition
A crucial question raised is what intermediates and products are responsible for the decay of the perovskite films caused by moisture.
To answer this question, MAPbI 3 films on various substrates are generally exposed to carefully controlled environmental conditions and exposure times. 11,79-81 Despite some existing discrepancies, an apparent conclusion is that water plays a catalytic role in the whole degradation process, thereby speeding up the decomposition of perovskite. A conclusive MAPbI 3 degradation reaction solely affected by water is shown in eqn (8).
Nevertheless, a large number of degradation routes and decomposed products, summarised in Table 1, have been proposed in each specific circumstance and some are quite different, even just slightly modulating the exposed conditions. Firstly, for identifying the intrinsic stability of perovskite, MAPbX 3 is either placed in a dry N 2 atmosphere or under vacuum conditions. In this duration, perovskite films have a negligible degradation except with other treatments like light illumination and high temperature, 82,83 consistent with the recent acknowledgement that light-activated oxygen degradation and thermal effects have substantially key roles in the degradation process. [84][85][86] Deretzis et al. proposed two possible degradation mechanisms for the degradation under air and vacuum conditions (Table 2): 87 (1) chemical degradation involving the catalyst role of water; (2) thermodynamic degradation through the creation of volatile molecular defects even partially occurring under vacuum. Following this work, Alberti et al. found that CH 3 NH 3 PbI 3 goes through a similar degradation dynamics pathway, beginning with a phase change from tetragonal to  cubic phase with no hydration taking place. 88 Water was just suggested to speed up the intrinsic thermodynamic mechanism by transferring the proton of MA + to I À and releasing CH 3 NH 2 . Furthermore, several investigations have presented the performance of perovskite materials in an artificially sealed humid environment. Christians et al. found sole formation of MA 4 PbI 6 Á2H 2 O without detecting PbI 2 following exposure to an artificially airtight humid atmosphere created by a water/glycerol solution. 89 Likewise, Zhao et al. reported that MAPbI 3 powder was fully recovered if no loss of PbI 2 and MAI occurred after drying liquid water. 90 Moreover, a very recent study exhibited that an in situ humid atmosphere causes crystalline MAI to become an amorphous phase without breaking its bonding structue. 91 These groups all reported no loss of MAI following the sealed humid air exposure treatment. However, Lin et al. observed that in a sealed deuterium oxide (D 2 O) atmosphere, the decomposition of MAPbI 3 was initialized by vaporization of CH 3 NH 2 , accompanied by quick formation of the PbI 2 phase. 92 Fig. 9 provides a view of the ion distributions from 3D perspectives as exposure time increases. 92 Flowing moist inert gas also accelerates the hydration of perovskite and facilitates the irreversible decomposition towards PbI 2 via evident phase separations. 90 However, no matter whether perovskite films degrade in a sealed humid atmosphere or exposed to flowing moist inert gas, the decomposition rate of perovskite crystals is substantially related to the exposure degree towards R.H. 81 (Fig. 10). 93,94 Further irreversible decompositions present the separate appearance of porous and distorted hexagonal PbI 2 platelets in the film. 95,96 Given that the working PSCs in actual situations are contemporaneously under direct exposure to light, heat, water etc., the coupled effects of moisture with other environmental factors on MAPbI 3 should also be considered. Studies presented that the cooperative effects of water with light, ambient air and high temperature would aggravate the water-induced influence. 11,95,[97][98][99] In these cases, many groups revealed that deterioration of perovskite films could occur at lower R.H. with different degradation pathways and final products (Table 1). For instance, Shirayama et al. recently indicated that two competing ways, perovskite hydration and the formation of a PbI 2 phase, occurred spontaneously under the cooperative effect of ambient air flow (R.H. 40%) and light. 97 Niu et al. reported that humid air with oxidization of oxygen and light illumination induced decomposition of MAI by losing I in gaseous form or solid I 2 , 11 as did Dao et al., 100 partially due to the deprotonation effect of O 2 in the trapped charge-induced MAPbI 3 degradation mechanism. 101 Furthermore, the Pb solid remnant was found as PbO, PbCO 3 , Pb(OH) 2 and other forms (Table 1 and eqn (9)-(13)) when the decomposition involved CO 2 and O 2 . 95,102 Also, Ruess and colleagues found that the coupled effects of illumination and moisture led to segregation of the mixed halide to MAPbI 3 and MAPbBr 3 in MAPb(I 1Àx Br x ) 3 followed by the appearance of PbI 2 via decomposition of MAPbI 3 . 80 Very recently, a new phase PbI(OH) was reported due to the transformation of PbI 2 at the interface of the perovskite and Spiro-OMeTAD after continuous illumination and moisture exposure (R.H. 60%). 103 An additional solid remnant, amorphous C, is observed on Au coated Si wafers. 79,104 The appearance of this hydrocarbon layer may be due to the interaction of low-energy electrons with MAPbI 3 by initially triggering the breakdown of C-N bonds. 105 Besides, it is probably instead that the unique carbon solid comes from the catalyst of Au under light irradiation. Electrons gain their own energy which is enough to activate the MA + group. 106,107 Beyond the obscure solid remnant, the probing of MAPbI 3 decomposition involves puzzling gaseous products. The existence of gaseous products has been confirmed by the bubbles at the interface of the Al electrode, 108 whereas gaseous components deriving from CH 3 NH 3 I are not unveiled yet and speculated to be diverse species, e.g. I 2 , HI, CH 2 NH 2 , NH 3 , and H 2 . 11,79,95 In future work, characterization and analyses of the gaseous products are expected to pave the way for overall understanding of the perovskite decompositions.

Effect of H 2 O on perovskite structure
The ingression of water is found to initially chemisorb on a particular MAPbI 3 region, extracting electrons from the perovskite surface like a p-type dopant. 79,109 Grain boundaries substantially  Fig. 11, while new grain boundaries are generated in this duration. 93 Chiang et al. proposed that pinholes and defects produced by light and heat within the grain boundaries could explain the phenomenon that large grain size is conducive to the long-term stability with or without moisture. 112 At the same time, Wang et al. found that an amorphous perovskite region exists between the perovskite grains, responsible for the quick water ingression parallel to the substrate, further emphasizing the significance of increasing grain size and improving grain boundary qualities. 113 In addition, trapped charge along grain boundaries is demonstrated to stimulate the irreversible deterioration in moisture-induced degradation. 101 A local electric field is assumed to deprotonate the organic cations by transferring the proton onto neutral H 2 O molecule, releasing volatile compound CH 3 NH 2 or HNQCHNH 2 , similar to the early acid-base mechanism proposed by Frost et al. 114 Through investigating the crystal surfaces of as-grown and degraded MAPbBr 3 perovskite single crystals, Murali et al. revealed that high humidity would restructure the crystal surface by surface hydration, yielding polycrystalline structures and grain boundaries. 115 Grancini et al. reported that the perovskite single crystal edge is more sensitive to moisture, where hydrogen bonding occurs spontaneously between water molecules and the perovskite lattice, assumed as the first step of the hydration process. 116 Further permeation of water into the inner structure of the perovskite is proposed to be related to the perovskite lattice, also determined by their different terminations (MAI-terminated and PbI-terminated) and the polarities. 110,117,118 The immigration can occur very fast within a few seconds at low R.H. with no obvious change appearing in optical absorption and structure. 119 In the (001) surface, the MAX-termination, although structurally more stable than the PbX 2 -termination, 120,121 presents B0.3 eV absorption energy of water with the large inner structural space, implying that water molecules can move freely with almost no constraint. 118 The MAI-terminated surface may experience a quick decomposition called a solvation process. 110 Iodide atoms, as shown in Fig. 12A, are replaced by water molecules, bonding with the MA molecules and escaping in the form of MAI molecules, 110 similar to the super alkali halide crystal model proposed by Fang et al. 122 In the case of a PbI 2 -terminated (001) surface, no solvation process happened due to the shorter and stronger Pb-I bonds. 110 Instead, water molecules directly went into the perovskite crystal structure due to the decreased energy in the diffusion and large interspace in the lattice structure, finally taking up a position in the perovskite slab (Fig. 12B). Another simulation recently indicated that the existence of O 2 could induce higher H 2 O molecules adsorption on the PbI 2 -terminated surface, further confirming the synergistic effects of H 2 O and O 2 . 99 Furthermore, water adsorption is also investigated on other perovskite surfaces. Zhang et al. revealed that in the (110) surface, hydroxyl groups and hydroxyl radicals  rather than H 2 O molecules are responsible for the quick deprotonation of MA, 102,123 in accordance with an ordinarily fast decomposition caused by the remaining hydroxyl radicals and hydroxyl groups in contact with the ZnO/perovskite interface. 124 Lv and colleagues showed that the erosion of water on single crystal MAPbI 3 is slower both in (100) and (112) facets but not in the (001) facet, attributing the different erosion rates to local atom arrangements. 125 Humidity-induced phase transitions in many studies include the formation of perovskite hydration compounds (MAPbI 3 ÁH 2 O and MA 4 PbI 6 Á2H 2 O) with the structure homogenously changing from 3D MAPbI 3 crystals to MAPbI 3 ÁH 2 O, a 1D structure and MA 4 PbI 6 Á2H 2 O, a 0D network (Fig. 13), 94,126 with the column of the perovskite unit cells swelling, 94,127 which may lead to lattice stress and eventually polycrystalline films. Detailed analysis manifested that MAPbI 3 crystals experience a space group transformation, derived from a change of hydrogen bonds between the MA + cation and I atoms after H 2 O insertion. 128 Likewise, new hydrogen bonds are proposed to be established between the O of H 2 O and H in NH 3 as well as between H in water and halide anions. 127 It should be noted that not all reports confirmed the generation of MA 4 PbI 6 Á2H 2 O, 90,103 which may be due to the different perovskite phase, a phase and b phase, with change of the I-Pb-I bond angle. 103 Although the back reaction can occur under low R.H. conditions, structural changes including destruction of the original long range order and porous microcrystal solid, caused by dihydration, are still commonly found in the microstructure. 93,[129][130][131] The decomposition of the perovskite is accompanied by a change in film thickness. 94,97

Effect of H 2 O on perovskite properties
Keeping in mind that the water molecules chemisorbed on the film surface extract electrons, the valence band maximum (VBM) and conduction band minimum (CBM) of the MAPbI 3 are shifted towards lower values. 79 Further evidence is the declining resistance of the MAPbI 3 films following the power law distribution in moisture, proposed to exhibit p-type conduction. 109 Nonetheless, chemisorbed water does not evidently change the bandgap (E g ), partially because the water molecules do not chemically react with the perovskite surface but interact electrostatically. 79 In addition, water molecules on the perovskite surface were revealed to reduce deep electron traps as well as to slow down the recombination rate. 132 But more substantial lattice distortion in the crystal edge is also reported, widening the local bandgap along with faster carrier recombination. 116 Another study showed that aged single crystal surfaces and polycrystalline films both obtain similar carrier lifetime and band gap, but poorer than crystal bulk. 115 Lately, lattice surface hydration has been found to improve the local bandgap of MAI À exposed crystals by a very small value by the slight upshift of the CBM, whereas water expands the bandgap of PbI 2 -terminated slabs by B0.30 eV. 110,118 Also, the optical proprieties are unchanged when the water absorbs on the MAI À surface. However, after H 2 O pierced into the inner space, the local bandgap increased from B1.6 eV to 3.1 eV (hydrated perovskite) and finally became 2.4 eV (PbI 2 remnant). 83,94 Another indication of the decomposition of MAPbI 3 is the striking color change from dark brown to yellowish, in agreement with the remaining absorption o525 nm. 11,95,96 The hydrated layers are proposed as insulators to block the transport of carriers and thereby instigate more recombination of electrons or holes. 94 Recently, Long et al. reported that the acceleration of electron-hole recombination can be attributed to the high-frequency polar vibrations of water. 132 Irreversible decomposed product PbI 2 caused by water exposure is found to spatially contaminate the conductivity within the grain inner regions and grain boundaries. 63

H 2 O at interfaces between perovskite films and functional layers
Thus far, we have concluded how water impacts intrinsic perovskite films. Furthermore, another key factor, interfaces between perovskite and other functional layers, has been recently pointed out to immeasurably affect the device lifetime. 89,108,[133][134][135] To investigate the relationship between interfaces and moisture, the FFL and SFL will be paid special attention. The nanostructure and nature of the FFL play unignorable roles in the water-induced interfacial degradation (Fig. 14). 131,136 The electron selective TiO 2 layer in n-i-p type cells may improve moisture resistance via helping the perovskite to better crystallize, 131 and further change the degradation degree by manipulating its intrinsic nanostructure. 136 In contrast, electron selective material ZnO in n-i-p cells, hole transport material PEDOT:PSS in p-i-n cells and insulating material Al 2 O 3 define themselves with hydrophilic qualities as accomplices in the intrusion of moisture. [136][137][138] The increase of the broad peak of O-H stretching in FTIR after exposure to water shows that the hygroscopic Al 2 O 3 film induces the water in air to ingress in the form of droplets. 136 IR also reveals that an acid-base reaction occurs at the contact of ZnO with NH 3 + groups of methylammonium cations. 136 Hydroxyl groups and acetate ligands are often kept on the ZnO particle surfaces and accelerate the transformation of MAPbI 3 to PbI 2 especially when the ZnO/perovskite structure is kept in a high temperature, consistent with theoretical simulations (Fig. 15). 123,124,139 Besides, another reason could be the more basic surfaces of ZnO compared with those of TiO 2 or ITO, which would facilitate protons to transfer and produce methylamine. 124,137 Also, an aqueous solution of ZnI 2 was proposed by Dkhissi et al. to  probably exist in this degradation. 137 But in one case, a small amount of PbI 2 at the perovskite surface and interface with ZnO, derived from high humidity treatment, was proposed to remarkably slow down the degradation of PSC performance, and contribute to the difference in chemical kinetics. 75 Capping functional layers and their contact interface with perovskite are other aspects that impact the moisture-induced aging (Fig. 14). Although Li-TFSI doped Spiro-OMeTAD is wellknown for its hydrophilicity, 81,140 few studies explored the detailed reason for this until Song et al. recently proposed that the moisture-invoked degradation at the interface is inhomogeneous, and assumed that the spontaneous existence of severe hydration and repression of recombination centers was induced by uneven distributed H 2 O and MAI in the Spiro-OMeTAD film. 131 Besides Spiro-OMeTAD, in situ experiments showed that PCBM in inverted PSCs absorbed water and deteriorated perovskite from the interface. 108,141,142 In a recent investigation, Zhao and colleagues directly contacted an Al electrode with perovskite and found that the first step for MAPbI 3 degradation is the redox reactions at the interface proceeded by moisturedriving ion migrations. 143 They also pointed out the importance of seeing the PSCs as a whole for discussing stability issues. Also, the Au electrode was used to slow down the release rate of trapped H 2 O molecules, based on the longer remaining time of the dihydrate phase. 144

Influence of H 2 O on PSC performance and stability
The influence of H 2 O on perovskite films during processing and the underlying mechanisms were discussed with further discussion in Section 2. A certain amount of water can improve perovskite film quality in most fabrication cases, which could also enhance the PSC performance. We have summarized this in Table 2.
The power conversion efficiency (PCE) is an important parameter to evaluate the initial performance of solar cells, which depends on the short-circuit current J sc , open-circuit voltage V oc and fill factor FF. J sc is generally determined by the bandgap, light absorption and carrier lifetime/diffusion length. The bandgap is an intrinsic material property and irrelevant to perovskite film quality. More surface coverage and thicker active layer can increase the light absorption. It is reported that H 2 O in the solution fabrication enhances the efficiency of light absorption, 33,40,76 partially due to improved coverage. 35,76 Nevertheless, ambient water vapor in the spin-coating step in OSSD obviously reduced the coverage of the CH 3 NH 3 PbI 3 film, thus leading to a poor J sc originating from the high transmittance and low absorption of light. 43 In some cases, the carrier diffusion length and carrier lifetime is long, contributing to efficient carrier collection, 54,59 which could be considered as the result of smooth surface, 33 better crystallinity, 76 and reduced grain boundaries. 48 V oc mainly depends on the recombination pathways, especially nonradiative recombination centers in the solar cell. The trap states are often generated by poor quality of the perovskite films. Water in synthetic stages could reduce the trap states in boundaries, 31,40,54 and promote crystallinity 54 to ensure higher V oc . This is also supported by the reduction of the nonradiative recombination loss. 33,54,59 Wang et al. employed water molecules to catalyze the reaction of MAI and PbI 2 after spincoating in SSD. 54 The V oc was 13.6% higher than that of thermal annealing processed PSCs, referring to higher crystallinity and less boundaries. However, Clegg et al. attributed the loss of open circuit voltage to the crack and pinholes after adding water to the PbI 2 DMF. 42 Nonradiative recombination traps contribute to the FF value as well. Larger crystal grains 39,40,48 39 The grain size significantly increased from 50-500 nm to 500-1000 nm, equal to the thickness of the film. Also, the incorporation of water is reported to enhance the surface coverage, reducing voids and pinholes significantly when annealing in B35% R.H. 57 However, it is not always true that the presence of H 2 O has a positive effect. When Coning and coworkers increased the water content in CH 3 NH 3 PbI 3Àx Cl x precursor solution, larger pinholes were observed in SEM. They likely provided direct contact between TiO 2 and P3HT, responsible for decreasing the fill factor to 0.67. 36 The PSC device stability under humidity can also be evaluated by modeling the change of performance parameters with time 145,146 or relative humidity. 147 Darvishzadeh et al. modeled the change of V oc with time, showing that the ion migration and PbI 2 defects are the major reasons for degradation of V oc upon moisture exposure. 145 Sohrabpoor et al. developed models, which showed that the FF is mainly responsible for the PSC degradation. 146 Both concluded that the defect profile played a key role in the device instability. 145,146 Moreover, a drift-diffusion based model and a two-diode model were offered by Bhatt et al. to explain the variation of device optoelectronic and electronic parameters, e.g. the absorption coefficient and diffusion length as well as their effects on the device performance with different relative humidity levels. 147 Besides the standard device parameters, the hysteresis of the J-V curve during the PSC measurement 148 was also significantly affected by water. Two kinds of causes for hysteresis have mainly been proposed: (1) ionic migration; (2) surface and interface defect state trapping or detrapping charges. 149,150 Mobile ions including MA + , 149 I À , 151,152 etc. migrate in the bulk point defects and the surface and grain boundaries as illustrated in Fig. 16. 153 Moderate water contents in the film preparation could reduce the hysteresis. Wu et al. added a moderate amount of water to the PbI 2 DMF precursor, fabricating a well crystalline, smooth perovskite film without pinholes, detecting no hysteresis. 39 Similar suppression was observed by Gong et al. when adding water into the MAPbI 3Àx Cl x precursors. 35 Gangishetty et al. found that the perovskite film prepared in almost 0% water exhibits severe hysteresis while the degree of hysteresis could substantially decrease as the humidity increases. 48 They attributed this to the reduced grain boundaries and traps. 48 In all of these cases, large crystal sizes and high quality films were observed with less boundaries and defects existing. This implies that less ionic channels, less defects for trapping and detrapping charges and better contact with other functional layers can be achieved. 39,48,150 Several authors have reported that the hysteresis becomes severe when placing the PSCs under a moist atmosphere. 89,94,95 For instance, Huang et al. reported that the hysteresis index increased from 0.14 AE 0.06 to 0.43 AE 0.23 after 7 days of exposure. 95 While the ingressive H 2 O interacts with Pb 2+ groups to form monohydrate MAPbI 3 ÁH 2 O as well as dihydrate MA 4 PbI 6 ÁH 2 O, it results in lattice expansion. 94,127 Thus the bonding between MA + and PbI 6 4À becomes weak, so the activity energy for ionic migration, which is seen as the hopping process, 154 also decreases, and thereby the migration of MA + and I À becomes obvious, as supported by the observations of fast degradation under an applied electronic field with moisture. 155 Therefore, increasing ionic migration leads to a severe hysteresis after H 2 O ingression. In addition to promoting migrations of ions, the moisture may induce exacerbation of hysteresis via increasing the amount of surface states at the surfaces and interfaces. 95,156 In the previous section, we have mentioned that the ingression of H 2 O molecules destroys a whole perovskite grain into small parts and increases the number of grain boundaries. Therefore, more defect states in the surfaces and interfaces can be expected. Snaith et al. have proposed that the defects trap and detrap charges, 148 which is consistent with the observations in experiments. 101,157 Moreover, experiments used KPFM and AFM to prove that the grain boundaries pave the way for the motion of ions such as FA + , and MA + , acting like a fast transporting channel, 158 and ultimately aggravating hysteresis. 156 To further understand the PSC stability under moisture, we summarize the performance degradation of several representative PSCs in terms of PCE in Table 3. Regardless of perovskite compositions and device architectures, PSCs undergo degradation under a humid atmosphere. However, it is still hard to make a conclusion about which kind of PSC obtains longer lifetime under exposure to H 2 O since substrates, environmental factors, etc. all contribute to the lifetime of PSCs.

Recent strategies to enhance stability
Given the degradation mechanism of perovskite materials and PSCs under water, the resistance of perovskite materials to water is relevant to intrinsic perovskite characteristics and the extrinsic factors including active functional layers, and device structure. To address the issues of aquatic erosion and to sustain the longevity of the perovskite, three basic principles could be considered. Firstly, the key intrinsic improvement is to enhance the quality and stability of perovskite single crystals and perovskite films. Secondly, to avoid decay of performance, the ingression of water molecules into the perovskite needs to be inhibited or prohibited. Thirdly, to guarantee recycling of water-damaged perovskite layers, healing strategies for the cell need to be developed. As discussed above, the ingression of water molecules starts from the grain defects and then diffuses into the inner structure through the boundaries. The hydrogen bonds between stagnated H 2 O and perovskite units release I atoms from the adjacent Pb 2+ , which in turn react with the loose MA + . DFT shows that the energy of MAPbI 3 is 0.1 eV lower than the total energy of free MAI and PbI 2 phases, indicating that a slight disturbance from water molecules could induce the decomposition of MAPbI 3 units. 159 Thus, improving the resistance is linked to the decrease of potential infiltration sites and the strengthening of structural bonding. 122,160 6.1.1 Modification of perovskite components. The tenability of RAPbX 3 perovskite guarantees multiple ions of RA and X to improve the stability of perovskite units. Anion X, referring to I À generally, has been partially replaced by Br À or Cl À , due to the stronger bonding and shorter perovskite bond length. 5,18,[160][161][162] Besides mixed halide anions, the SCN À anion is appealing to many researchers due to its large formation constant with Pb 2+ , 163 and RAPbI 3Àx SCN x (RA = MA + or FA + ) films show less defects, better crystallinity and pronounced resistance towards moisture infiltration. [163][164][165] Furthermore, SCN À additive in FAPbI 3 has been proven to inhibit the formation of the d-FAPbI 3 phase as well as facilitate the growth of the stable a phase which eventually offers SCN À -doped FAPbI 3 PSCs better moisture stability. 165,166 Satyawan et al. proposed another latent anion BF 4 À that could partially substitute I À to fabricate MAPbI 3Àx (BF 4 ) x with similar ionic radius, 167 and lately MAPbI 3Àx (BF 4 ) x has been employed in a HTL-free cell architecture by Chen and coworkers, presenting improved stability due to the hydrophobic F atom. 168 Completely or partially using inorganic cations e.g. Cs + , 169-172 2D organic cations like small-molecule bulky ammoniums (e.g. C 6 H 5 (CH 2 ) 2 NH 2 + , 173-175 CH 3 (CH 2 ) 3 NH 3 + , 176,177 ethylenediamine 178 ) and polymer cations (e.g. PEI 179,180 ) for the A site are of interest. 2D organic cations equip perovskite films with high stability under H 2 O because of the van der Waals force in the perovskite layers making the structure robust and the hydrophobicity of 2D organic cations. 173,174,176 Nevertheless, 2D organic cations have been known for a long time to endanger the efficiency of PSCs due to the inhibition of the organic component on the carrier transport, which may hinder the passion to explore the potential of 2D or even lower dimensional organic cations. 176,181,182 Luckily a very recent study showed that the ebb of the efficiency could be mitigated by pre-heating the substrate while maintaining the resistance towards moisture. 183 Fig. 17 shows that (BA) 2 (MA) 3 Pb 4 I 13 (BA = butylamine) exhibits a close singlecrystalline quality with the inorganic perovskite component vertical to the substrate. 183 This significantly improves the carrier mobility and reduces the recombination of free charge carriers, bringing the efficiency up to 12.52% with no hysteresis and even no decrease in efficiency after encapsulation for 42000 h.
6.1.2 Fabrication of high-quality perovskite films. Polycrystalline perovskite is the primary form for perovskite studies in the optoelectronic and photovoltaic fields. Nevertheless, a general polycrystalline film consists of many small grains, accompanied by grain boundaries and interfacial defects as we discussed above. All of these play impellent roles in the penetration of H 2 O. Accordingly, to improve the device stability while maintaining high efficiency, an essential demand is to synthesize a perfect polycrystalline film without remnant trapped hydrophilic molecules. In addition to carefully controlling the relative humidity at its optimal value and the exposure time in each fabrication step, studies try to adjust other process parameters and conditions of solution-based or vapor-based techniques. 29,72,[184][185][186][187][188][189][190][191][192][193][194][195][196] Besides, traditional ceramic or polymeric membrane fabrication techniques like screen printing, 197 bar coating, 198 etc., 60 are employed in PSCs.
Monocrystalline perovskites have attracted lots of interest because of their pronounced electronic properties and their role in mechanisms studies. 199,200 The impeccable crystallinity and stability ensure that perovskite single crystals are a remarkable candidate for photovoltaics. 201 This has been confirmed via a device structure of ITO/monocrystalline MAPbBr 3 /Au showing stable output in moist air after exposure towards R.H. 55% for over 48 h, while that of a polycrystalline MAPbBr 3 based structure drops to B60% of its initial level. 202 However, the best reported efficiency for a monocrystalline perovskite photovoltaic device is about 6%, still far below that of the certificated In addition, considering that MAPbI 3 is fabricated in working devices, the impact of interfaces and substrates can offer another effective route for enhancing the intrinsic stability of perovskite bulk films. 205,206 Recently, F4TCNQ acted as an interfacial layer in Spiro-OMeTAD/perovskite, reduced the trap states of modified perovskite by surface passivation and interfacial doping and hence improved the long-term stability. 206

Protection of perovskite from ingression of H 2 O
Although modification of the perovskite itself has a striking feature in the improvement of resistance towards moisture, most of the stability tests are still far from satisfying the standard moisture endurance requirements. To further improve the moisture tolerance, a number of studies have focused on isolating perovskite from H 2 O from the view of solar cell architectures. Here, we start from the FFL to the electrode based on the device structure, and conclude with how the functional layers and device structure present their features in retarding and prohibiting the invasion of H 2 O.
6.2.1 Optimization of the FFL. As we mentioned above, one of the penetration pathways for moisture is migration through the interface between MAPbI 3 and the FFL, including ZnO 2 , TiO 2 , PEDOT:PSS etc (Fig. 14). 207 The hydroxyl groups on the surface of ZnO in n-i-p cells and the hydrophilic and acidic properties of the hole transport layer PEDOT:PSS as the FFL in inverted p-i-n cells both accelerate the decomposition of CH 3 NH 3 PbI 3 from the contact interface. Thus, a better FFL and its interfaces with halide perovskite are required. One strategy is to further optimize the chemical properties of TiO 2 , ZnO and PEDOT:PSS. [208][209][210] It is also noted that some studies have attempted to insert a layer into the interface, aiming to separate the perovskite from ZnO. [211][212][213] However, very few studies are realized via these optimization approaches. Instead, studies are more inclined to design new materials, classified as metal-based, 214-216 metal oxide-based 217-219 and carbon-based ones, 220-222 featuring hydrophobicity, non-corrosion and a superior contact with the organic-inorganic hybrid halide perovskite.
6.2.2 Hydrophobic treatment of perovskite films. In addition to enhancing the perovskite intrinsic stability in the presence of water as we concluded in the above section, a few studies have tried to modify the perovskite film with hydrophobic treatments. Small molecules or ionic molecules with hydrophobic groups are utilized to cover the perovskite surface, as in-site water-repulsed layers. 191,[223][224][225][226][227][228] For instance, Yang and colleagues demonstrated the direct synthesis of a hydrophobic anion layer like tetraethyl ammonium (TEA) onto the perovskite film by dipping the as-formed MAPbI 3 film in isopropanol solution. 226 The TEA anion layer hardly impacts the efficiency compared with the control group, and it also maintained the majority of the original efficiency after storage in B55% R.H. for 500 h. 226 Furthermore, an unprecedented concept of a ''polymer/perovskite composite'' was proposed in 2014, 229 featuring the isolation effect on H 2 O molecules and the reduction of grain boundaries. 84,[230][231][232][233][234][235] One typical example is that Zhao et al. introduced a PEG 3D network into the perovskite films. 231 Strikingly, PEG absorbs water molecules and stops them from reaching the perovskite grains. Further work replaced PEG with a PCBM/PEG composite to better suppress hysteresis. 233 6.2.3 Optimization of the SFL. Spiro-OMeTAD as a SFL in an n-i-p device structure presents unsatisfactory film-forming qualities and often hydroscopic additives like Li-TFSI are employed, which go against the desire for long-term running under the operating conditions. Therefore, to alleviate the damage from water to some extent, efforts have been devoted to optimizing Spiro-OMeTAD hole transporting materials by fabricating high-quality films 236 and creating new dopants. [237][238][239][240][241] Another approach involving the removal of Spiro-OMeTAD is to design waterproof materials. [242][243][244][245][246][247][248] One of the prominent examples of designing new materials was that Zhang and coworkers applied an inorganic material CuGaO 2 as the HTL in the n-i-p device architecture, yielding over 18% efficiency which is comparable to that with Spiro-OMeTAD. 244 At the same time, the CuGaO 2based device almost retained its high performance after 30 d of exposure to moisture (R.H. = 30-55%) while the PCE of the Spiro-OMeTAD based device went down to 3% after a week.
For the p-i-n type PSCs, the SFL is generally the electron transporting material PC 61 BM. However, PC 61 BM was shown to interact with water and easily provided pathways for the direct contact of the Al or Ag electrode with the perovskite film, all of which damages the stability of p-i-n PSCs. 141 Some studies tried to solve this problem by doping PC 61 BM. 249,250 Another pathway is to employ other kinds of materials as the ETL in p-i-n type devices, like inorganic metal oxides, 141,251,252 or other organic mateirals. 250,253,254 The longevity of PSCs could be further improved by an appropriate choice of interfacial layers. 11,252,[255][256][257][258][259][260][261][262] Organic or inorganic interfacial layers lying in the electrode/SFL interface or in the SFL/perovskite interface, modified the moisture-sensitive interfaces and protected perovskite films from moisture infiltration and the degraded electrode. For instance, Niu et al. inserted an inert Al 2 O 3 layer into the interface between Spiro-OMeTAD and MAPbI 3 , enabling PSCs to better resist moisture at room temperature while retarding the recombination of the charges at the interface. 11 6.2.4 Optimization of electrodes. One primary pathway for moisture infiltration is going through the electrode, and the SFL and then penetrating the inner structure of the PSCs, thus the electrode works as a forerunner to stop H 2 O molecules. Two common electrode materials, Ag and Al, although low cost and possess good electron transport ability, have a poor chemical stability, not suitable for long-term operation. 143,263 A noble metal electrode Au, has shown better moisture resistance, but the high cost limits it in the large-area fabrication and a recent study reported that it could trap water molecules 144 and was able to diffuse through the hole transporting layer. 264 More recently, Huang et al. presented the very appealing potential of Cu electrodes in PSCs. 265,266 The unencapsulated PSCs employing Cu electrodes could maintain almost 100% of the original efficiency value after being placed in B55% R.H. at room temperature for over 800 hours. 266 However, Cu metal undergoes oxidation to Cu(OH) 2 and/or CuO x after exposure to oxygen or moisture, and thus it still requires strict encapsulation for practical applications. To date, a widely-accepted hydrophobic material is the carbon family including graphite, carbon black, mesoporous carbon and some graphene-derived nanomaterials. 267 This family not only provides chemical stability, substantially increasing the hydrophobicity of PSCs, but also maintains good conductivity with low cost. Han et al. successfully developed a typical model based on a triple-layer structure TiO 2 /ZrO 2 /MAPbI 3 with 5-AVA additive. 197 The unchanged output in 1000 h is ascribed to the stable contact of TiO 2 with perovskite and the hydrophilic effect of the porous carbon electrode which blocks upper moisture completely. 197,207 Further studies improved its efficiency up to 15%, making it more competitive compared with the highly performing metal electrode-based PSCs. 268 6.2.5 Device encapsulation. Given the success of the device encapsulation technique in organic light emitting diodes, dye-sensitizer solar cells and polymer solar cells, it has also been applied in PSCs to isolate H 2 O chemically at the molecular level or to block H 2 O mechanically at the macroscopic level. Encapsulation has shown its simplicity by only spinning casting and drying the hydrophobic polymer Teflon on top of the electrode while maintaining superior hydrophobicity towards moisture. 269 Encapsulation stability also demonstrated its viability via outdoor tests operated, respectively, in Jeddah, Saudi Arabia 270 and Hong Kong. 271 These all indicated the plausible tolerance of encapsulation in PSCs towards natural environmental conditions. However, moisture still slowly penetrates inside through the edges of the adhesive layer or directly through the sealing material (Fig. 18), [272][273][274] showing immeasurable damage capacity especially at 85 1C. 272 The still limited longevity of encapsulated PSCs points out that the choice of sealing materials and the airtightness of the contact interfaces between the encapsulated materials require more attention in the future. Furthermore, Wilderspin et al., for the first time, proposed a very simple approach based on RGB analysis to evaluate the feasibility and effectiveness of encapsulation approaches. 275 However, to meet the demand of finding a more robust sealing method and evaluation of other techniques for improved stability, we believe that more feasible and highly effective assessment methods should be developed in the future.

Curing decomposed perovskite caused by H 2 O
Although much work has been done to improve the stability of perovskite films and the corresponding devices which we also emphasized above, the existence of damaged perovskite films seems to be inevitable during device preparation and utility, indicating the importance of healing lacerated perovskite films. Now, a few curing approaches have been discovered and adapted to recover the original performance, 100,231,[276][277][278][279] and some are especially for water-induced deterioration, presenting promising healing effects. 90,231,276 Zhao et al. have demonstrated that no loss of MAI played a key role in voluntary, complete reformation of CH 3 NH 3 PbI 3 powders after drying water. 90 In PSCs, a 3D polymer scaffold PEG was applied to provide the CH 3 NH 3 PbI 3 film with self-healing abilities. 231 The recovered films exhibit compatible output compared with the initial performance, while the light absorption and J sc get improved after this recovery. In addition, the recovery can also be realized by other approaches like light irradiation, low moisture exposure, and thermal annealing. 276 By providing extra energy, respectively, they could drive complete recovery from decomposed phases to the tetragonal perovskite phase in several hours, offering effective and feasible routes to retreive the spoiled goods in large-scale industrial production as well as sustain the working devices in outdoor environments. It is demonstrated that both LiTFSI and TBP cause the Spiro-OMeTAD layer to be fragile upon thermal annealing and light irradiation. [280][281][282][283] Therefore, these two healing strategies should be avoided with sensitive materials. In fact, considering that the PSCs suffer from thermal stress and irradiation during long-term operation, further development of other stable Spiro-OMeTAD additives or stable hole transporting materials is one of the critical issues for operation in a natural environment.

Conclusions and outlook
In this review, we have analyzed the main impacts of water in various stages of the lifecycle of the state-of-the-art lead-based PSCs. It was shown that moderate water amounts could facilitate the nucleation and crystallization of the perovskite, resulting in better perovskite film quality and increased PSC performance. However, water irreversibly destroys the perovskite materials after reaching a certain level, but they exhibit better tolerance against water than initially expected. Hence, humidity resistant fabrication of high-performance PSC devices and modules is required. Generally, water shows a negative effect on the longterm stability of PSCs. Self-healing behavior has been found in perovskite materials and the recovery of PSC performance can be realized by adopting appropriate measures. Research on elucidating the connection of perovskite materials and solar cells with H 2 O has made significant progress in the past few years. Nevertheless, there are still some basic questions and practical issues regarding this connection in the fabrication process and long-term operation, which require further considerations elaborated in the following. Protection methods should be adopted to reduce the water effect during the practical operation of a PSC.

Materials and device structure
The role of water in the fabrication part is often related to MAPbI 3 , but more knowledge on other doped perovskites is likewise needed, because recent high-performance PSCs contain Cl À /Br À or are doped with FA + . The controversies found in the literature indicate that the impact of water could also be attributed to other factors, e.g. device structures, which calls for a more systematic study of the affecting factors. To meet the trend of green solar cells, less-lead or lead-free perovskite materials should be developed. The present development work on lead-free perovskite materials such as Sn, Bi, Cu, and Sb based perovskites and derivatives as well as double perovskites often focus on intrinsic properties and optoelectronic performance, whereas their relationship to water is not adequately addressed. Hence, it would be beneficial to provide a reference value for water tolerance for practical cases.

Fabrication
Previous studies have mainly dealt with the impact of water in single fabrication steps, which enhance the understanding of the influence of H 2 O. However, studies which consider water or moisture in the whole synthetic process are very few. Better understanding of the acceptable range or steady optimized value of water amount in the whole process could help to reduce the complexity of the fabrication. In spin-coating processes, which are often bound to small-scale applications, the influences of water are well-known, whereas in the case of scalable industrial production such as screen printing, more research on the impact of water is still required. Concerning real applications of PSCs, the fabrication technique is of upmost importance to costs and volume meaning that identifying first the fabrication method could be useful followed by optimization of control of water in each stage, which may also necessitate lab studies.
A literature study showed that an appropriate amount of water in perovskite raw ingredients or precursor solutions has a positive effect, which suggests that water as an additive could improve the quality of perovskites. In the later fabrication steps, water mainly appears in the form of moisture during the film growth requiring precise control of the relative humidity. One feasible strategy to ensure correct the humidity ratio could be to use a gas mixture of water vapor and dry air in a given ratio and at thermal equilibrium, i.e. at the same temperature. When the humidity is difficult to control in the fabrication, an alternative pathway would be to develop water resistant fabrication methods based on compositionally modified or doped perovskites, or to use water absorbents which protect perovskite growth from moisture. The behavior of PSC modules under real conditions against H 2 O is not yet clear. The synergistic effect of H 2 O, light, and thermal stress on module size deserves more research in the near future.

Stability
Despite recent progress, the full basic understanding of the stability of the perovskite layer and its interfaces against moisture is still far from complete. For example, confirmation of gaseous decomposition products may require the use of isotope tracer techniques. The role of the interface in the moisture ingression is also unclear.
As the decomposition of perovskite films and devices seems to accelerate only after a certain amount of water, stability could be increased by improving the intrinsic perovskite film and optimizing device architectures. Furthermore, one could also expect that entirely eliminating H 2 O in the encapsulating step and improving the encapsulation techniques could enhance longterm operation in natural environments. This would also require stricter stability screening and considering the combined effects of water with other environmental factors when evaluating a new material or encapsulation technique. Following closer established procedures for environmental testing such as ISOS, IEC (61646), and long-term outdoor tests would be beneficial in this context. For PSC one could also consider introducing water vapor transmission rate (WVTR) and oxygen transmission rate (OTR) as additional test variables to be monitored, which are encapsulation quality concepts often used in organic light-emitting diodes and organic polymer solar cells.

Conflicts of interest
There are no conflicts to declare.