Instability of solution-processed perovskite films: origin and mitigation strategies

Perovskite solar cells (PSCs) are promising next-generation photovoltaics due to their unique optoelectronic properties and rapid rise in power conversion efficiency. However, the instability of perovskite materials and devices is a serious obstacle hindering technology commercialization. The quality of perovskite films, which is an important prerequisite for long-term stable PSCs, is determined by the quality of the precursor solution and the post-deposition treatment performed after perovskite formation. Herein, we review the origin of instability of solution-processed PSCs from the perspectives of the precursor solutions and the perovskite films. In addition, we summarize the recent strategies for improving the stability of the perovskite films. Finally, we pinpoint possible approaches to further advance their long-term stability.


Introduction
Metal halide perovskites have become one of the current research hotspots of energy conversion materials due to their superior optoelectronic properties, such as high carrier mobility [1], long carrier diffusion length [2], adjustable * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. band gap [3] and long carrier lifetime [4,5]. Besides, solution processibility enables perovskites to be coated by simple solution-based techniques with a high deposition rate of up to 180 m h −1 [6] and at a relatively low manufacturing cost. Solution-processed perovskites have been widely used in a variety of devices, including light-emitting diodes (LEDs) [7][8][9], photodetectors [10][11][12][13], x-ray detectors for imaging [14] and solar cells [15,16]. Their employment as absorbers in solar cells is extremely successful. The CH 3 NH 3 PbI 3 (MAPbI 3 ) perovskite was first used as a sensitizer in dyesensitized solar cells in 2009 with an efficiency of 3.8% [17]. Motivated by this pioneering study, a variety of solution-based methods (spin-coating [18], spray coating [19] and slot die coating [20], etc) and advanced strategies (e.g. anti-solvent

Future perspectives
Perovskite solar cells (PSCs) are promising next-generation photovoltaics due to their unique optoelectronic properties and rapid rise in power conversion efficiency (PCE). However, the instability of perovskite materials and devices is a serious obstacle hindering technology commercialization. Future research into the quality of precursor solution and the post-deposition treatment performed after perovskite formation will play a key role in the development of stable PSCs. For the perovskite precursor solution, designing novel solvent and reductants in perovskite precursors to inhibit the oxidization of I − is the key to enhancing uniformity and stability. For post-deposition treatment, composition regulation, strain engineering, defect passivation and phase stabilization could further enhance the stability after the formation of perovskite films. Further efforts can be made in developing regulation of the crystallization kinetics, novel functional layer with superior stability and advanced encapsulation methods to realize high performance and stable PSCs. [21], solvent annealing [22] and hot casting [23], etc) were developed to improve the quality of perovskite films. So far, the record power conversion efficiency (PCE) of solutionprocessed perovskite photovoltaics has achieved 25.8% (certified 25.5%) for small cell (∼0.1 cm 2 ) [24] and 22.72% for minimodule (24 cm 2 ) [25], which is much higher than other emerging solar cells [26] and is approaching single crystalline silicon cells [27]. When the comparison is made between solution-and vapor-based methods (e.g. thermal evaporation [28,29], chemical vapor deposition [30,31] and close-space sublimation, etc), solution-processed perovskite solar cells (PSCs) outperform vapor-processed PSCs because of better electronic quality (lower density of deep level defects) of perovskite absorber [32]. On the other hand, the stability of solution-processed PSCs is still not comparable to photovoltaics existing in the market (e.g. 20-25 years) and has become one of the most imperative challenges to be addressed before technology commercialization can be considered [33,34].
Perovskites are unstable under many external stimuli, such as moisture, oxygen, heat, light and reverse bias [35,36], which is a major obstacle hindering the long-term operational stability of PSCs [37][38][39]. In view of solution-processed perovskites, stability is closely related to the characteristics of the precursor solution and the quality of perovskite films. Purity and solubility of the solute, the interaction between solute and solvent, and the chemical reactions between different solutes may seriously affect the stability of the precursor solution and later on the quality of the perovskite films [40][41][42][43][44][45][46][47][48][49]. On the other hand, the chemical composition, defects chemistry and crystal phase of the as-prepared perovskite films determine their long-term stability. For example, the MAPbI 3 perovskite decomposes quickly under moisture or heat conditions via loss of the MA cation [50][51][52][53]. The FAPbI 3 perovskite demonstrates higher thermal stability but suffers from phase instability due to a mismatch of the ionic radius [39,54,55]. When multi-cations are introduced in the crystal lattice to form FAMA or CsFAMA perovskites, the phase instability of the FAPbI 3 can be mitigated. However, the stability of the complicated precursor solution and the thermal stability of the PSCs using these perovskites remains to be solved because of the presence of the MA cation [56][57][58]. Herein, we review the origin of instability of solution-processed PSCs mainly from the precursor solutions to the perovskite films (figure 1). In addition, we summarize the recent strategies for improving the stability of perovskite films. Finally, we pinpoint possible approaches to further advance their long-term stability.

Origin of instability in solution-processed perovskite
2.1. Perovskite precursor solution 2.1.1. Solute. PSCs can be fabricated by solution processing methods. Purity, solubility of the solute and the chemical reactions between different solutes in the perovskite precursor solution affect not only the reproducibility of fabrication processes but also the quality of perovskite films (e.g. the defect densities, composition, uniformity, etc), which determines the stability of PSCs. As a consequence, understanding the characteristics of the solute is the key prerequisite for pursuing long-term stable PSCs. Recently, the degradation mechanism of the solute in the perovskite precursor solution was systematically studied.
The FAPbI 3 powder serves as an excellent raw material for fabrication of high-performance FAPbI 3 devices [24,41,[59][60][61]. Shin et al investigated the effect of solutes on stability of the FAPbI 3 precursor solutions and the FAPbI 3 films [42]. Different precursor solutions were prepared either by dissolving a mixture of FAI/PbI 2 or the synthesized singlecrystalline α-FAPbI 3 in the same solvent (figure 2(a)). After aging the solution at ambient condition, the pH value of the conventional FAI/PbI 2 precursor solution was significantly decreased from 11 to 6 due to the formation of HI. The acidic condition prevents the formation of α-FAPbI 3 . On the other hand, the pH variation was less significant when singlecrystalline α-FAPbI 3 was used as solute. As a result, the PCE of PSCs prepared with the FAI/PbI 2 mixture decreased obviously with aging of the solution. While devices using α-FAPbI 3 based solute were quite stable at the same time scale. Chen et al found that the I − in the precursor solution was easily oxidized to I 2 . And the I − 3 appeared with the reaction of I 2 and I − [45]. After aging the FAI/MAI solution in air for 2 d, color of the solution changed to yellow. This indicated the formation of I − 3 , which deteriorated the performance and reproducibility of the PSCs. They improved the stability of the precursor solution by adding benzylhydrazine hydrochloride as a reductant, which could effectively reduce the I − 3 and stabilize the PSCs under operation condition.
Other degradation mechanisms of the precursor solutions were also proposed [47]. Chen et al found that the decomposition of FAI in the precursor solution could be divided into two steps, i.e. the deprotonation reaction and the additionelimination reaction (figures 2(b) and (c)). In brief, the FAI was firstly deprotonated to form FA and HI. Then three FA molecules were self-condensation to form s-triazine. The HI and s-triazine were the main decomposition product. Wang et al found that the MA and FA mixed cations were also not stable in the precursor (figure 2(d)) [56]. They probed the composition evolution of the precursor solution by using    Besides the solute, another major factor determining the characteristics of precursor solutions and quality of perovskite films (e.g. composition, crystallinity, morphology, and trap-state density) is the solvent. The solvents with different properties, such as boiling point, vapor pressure, viscosity, dipole moment and donor number (figure 3(a) and table 1), would not only affect the evaporation rate of the perovskite precursor solution but also the nucleation and crystal growth of the perovskite films [62]. The solvents with different solvating and coordination abilities play an important role in the formation of perovskite films. The affinity between PbI 2 and I − is higher than that of PbI 2 and solvent molecules with low solvation ability (e.g. acetonitrile (ACN), isopropanol (IPA) et al), which facilitates the formation of iodoplumbate complexes (PbI − 3 and PbI 2− 4 ) before the perovskite formation. On the other hand, high-affinity solvent molecules (e.g. DMPU, HMPA et al) coordinate with PbI 2 and suppress the crystallization of perovskite film via formation of the PbI 2 ·solvent complex. The introduction of solvents with different coordination abilities would change the equilibrium between PbI 2 ·MAI and PbI 2 ·solvent, leading to the transformation of coordination complex species and perovskite crystallization kinetics.
DMF and DMSO are the most conventional solvents for preparation of the perovskite precursor solution. These solvents show strong coordination ability and high boiling points. Xiao et al found that when DMF was used as the solvent, slow crystallization of the perovskite causing the formation of many uncovered pin-hole areas over the film (figures 3(b) and (c)) [64]. In addition, Yoo et al found that many pin-holes appeared in the perovskite films obtained from the precursor solution with a mixed solvent of DMF and DMSO (figures 3(d) and (e)) [63]. The balance between fast nucleation and slowed crystal growth was the key prerequisite for the formation of uniform and dense perovskite films. Crystal structures of semiconductors determine their electronic structures and optoelectronic characteristics. Zakutayev et al calculated the band structures of the III-V and II-VI semiconductors (e.g. GaN, GaAs, CdSe). The valence band maximum (VBM) of these semiconductors comprising the bonding states is easier to form deep trap states. These deep trap states can act as Shockley-Read-Hall recombination centers, significantly impeding carrier transport [65,66]. Different from these conventional semiconductors, organometal halide perovskite shows an ABX 3 crystal structure with a soft lattice (figure 4(a)) [66]. Yin et al calculated the band structure of the MAPbI 3 perovskite using density functional theory (DFT)-PBE ( figure 4(b)). They showed that MAPbI 3 is a direct bandgap semiconductor. The mixing of a Pb s orbital and an iodine p orbital constitutes an anti-bonding coupling in the VBM [37]. Meanwhile, the Pb p state contributes to the CBM. The unique band structure provides perovskite with better defect tolerance and extraordinary optoelectronic properties.
The FAPbI 3 , FAMAPbI 3 and CsFAMAPbI 3 perovskites with MA, FA and Cs as the A-site cations are commonly used to prepare high-efficiency PSCs. MA and FA-cation are unstable in the precursor solution which has been discussed in section 2.1. Besides, the stability of MA-and FA-based PSCs is also unsatisfactory, especially under humidity or heat condition. Understanding the cation-induced device failure mechanism is crucial for improving the long-term stability of PSCs.
The degradation of MA-contained perovskites under moisture or heat condition has been revealed to be the major instability pathway of such types of PSCs. Niu Gibbs free energy results indicated that the decomposition process can be accelerated by the presence of O 2 and under UV-light, see equations (2) and (3) [4HI (aq.) The instability of MAPbI 3 caused by humidity and oxygen could be solved after careful encapsulation. On the other hand, the heat-induced instability of MA-contained perovskite may be more severe. MAPbI 3 decomposes slowly at moderate temperature (65 • C-85 • C) and rapidly at high temperature (135 • C-150 • C) [35]. Conings et al compared the thermal stability of MAPbI 3 films under different atmospheres (figure 4(c)) [68]. After storage in N 2 atmosphere at 85 • C for 24 h, the electrical conductivity of the perovskite film is slightly decreased, indicating the beginning of the decomposition process. The decomposition process is accelerated in O 2 atmosphere. When heating in air, the decomposition rate of the film is greatly accelerated and the conductivity is greatly reduced.
Juarez-Perez et al used TG-DTA to study the thermal decomposition behavior of MAPbI 3 in He atmosphere [50]. There are two mass loss steps in the decomposition process of MAPbI 3 , which can be attributed to the loss of MAI (290 • C) and PbI 2 (above 420 • C) (figure 4(d)). Thermal decomposition of MAI follows different pathways at different temperatures, i.e. equation (4) at low temperature and equation (5) at high temperature. In addition, MA transmethylation reactions occur, as described in equations (6) and (7). The main decomposition product of MAI is CH 3 I [CH 3 NH 3 + (g) + I − (g) → CH 3 NH 2 (g) + HI (g) ] (4) Later on, Juarez-Perez et al investigated the thermal decomposition of FAPbI 3 by using the same technique [69]. Two mass loss steps are observed, which can be assigned to the decomposition of FAI (330 • C) and PbI 2 (400 • C) (figure 4(e)). The increased decomposition temperature of FAI in FAPbI 3 compared to MAI in MAPbI 3 indicates higher thermal stability. FAI mainly undergoes the following three thermal decomposition reactions. Equation (8) is the FA selfcondensation reaction at a low temperature. Equation (9) is the FA decomposition reaction at a high temperature. And equation (10) is the deprotonation reaction of FAI Although FAPbI 3 has higher thermal stability, it suffers from poor phase stability, especially under humidity condition, which will be discussed in section 2.2.4.
Shi et al used gas chromatography-mass spectrometry to study the thermal decomposition products of FAI, MAI and MABr powders [53]. They found that both FAI and MAI decomposed significantly in the temperature range of 85 • C-350 • C. After heating at 85 • C for 100 h, the decomposition products of MAI and MABr were CH 3 I and CH 3 Br. And the decomposition products of FAI were H 3 C 3 N 3 and NH 3 . After aging at 140 • C for 10 h, the additional decomposition product of MAI and MABr was NH 3 . After heating at 350 • C for 15 min, the decomposition products of MAI and MABr were the same as heating at 140 • C. While the decomposition products of FAI changed from H 3 C 3 N 3 to HCN. These results were consistent with the thermal decomposition reactions raised by Juarez-Perez et al [50,69].

Strain.
Strain in perovskite films is the main origin of instability and cannot be solved by the conventional extrinsic stabilization methods [70]. There are two types of strain in the PSCs, i.e. the local lattice strain and the external conditioninduced strain [38]. Saidaminov et al reported that local lattice strain in perovskite films arises from the ionic size mismatch between the FA-cation and Pb-I cage. The local strain contributes to the cage distortions and BX6 octahedra tilting, facilitates the formation of vacancies and results in the degradation of PSCs. (figure 5(a)) [71]. Zhu et al revealed gradient evolution of residual strain in the vertical direction of the mixed halide perovskite film [72]. They performed cross-section TEM and nano-beam electron diffraction (NBED) measurements on three typical regions at different depths ( figure 5(b)). The NBED patterns indicate the lattice distortion in the microscopic crystal structure, the increase in crystal plane distance and the decrease in lattice constant from the surface to the bottom.
External condition-induced strain may arise from the mismatch of thermal expansion coefficient between the perovskite films and the substrates. Xue et al summarized the thermal expansion coefficient (α) of each functional layer (figures 5(c)-(f)) [73]. They found that perovskites have much higher α values (ranging from 3.3 to 8.4 × 10 −5 K −1 ) than the ITO-coated glass and metal oxide charge-transport layers (in the range of 0.37-1 × 10 −5 K −1 ). A high-temperature annealing process (>100 • C) is necessary to allow crystallization of the perovskite films. This result in the formation of a tensile strain during the cooling process. They also calculated the stress in different perovskite films. Stress in the CsPbI 2 Br film is much larger than in the MAPbI 3 film because of a higher annealing temperature and a larger thermal expansion coefficient mismatch relative to the substrate [74]. In addition, mismatch of lattice constant between the perovskite and the substrate is another source of external condition-induced strain. Chen et al reported the strained epitaxial growth of α-FAPbI 3 thin films on lattice-mismatched halide perovskite substrates (figures 5(g) and (h)) [75]. The XRD peaks of MAPbCl x Br 3−x substrates shift to higher diffraction angles as the increase of x and the peak of FAPbI 3 shifts to the lower diffraction angles.

Defects.
Perovskite films usually have a polycrystalline nature with different types of defects. Although perovskites demonstrate high tolerance of defects, the performance and stability of the PSCs can still be affected especially in high-efficiency devices. From the performance perspective, the existence of defects affects the interfacial contact and the extraction of carriers, increasing series resistance and non-radiative recombination [76,77]. From the stability perspective, the defects (especially the bulk defects) accelerate ion migration during the operation of the device [78], causing the decomposition of the crystal structure [79] or phase separation [80,81]. Therefore, passivation of bulk, surface and interfacial defects is one of the effective methods to improve the efficiency and stability of the perovskite device.
Gao et al summarized the types of defects in perovskite films [82]. The typical defects in MAPbI 3 film could be divided into 0D and 2D defects (figure 6(a)). The Pb 2+ vacancies, I − vacancies, interstitial Pb 2+ , interstitial I − 3 and Pb-I anti-site substitution are considered as the 0D defects. While those at the grain boundary and on the surface are the 2D defects. Furthermore, Buin et al calculated the formation energies of the point defects in the MAPbI 3 films by using the DFT [83]. Several types of point defects were presented in figure 6(b), such as the vacancy defects (MA (V MA ), Pb (V Pb ) and I (V I )), interstitial defects (MA i , Pb i , I i ) and substitution defects (MA Pb , Pb MA , MA I , Pb I , I MA , I Pb ). The  [84]. They found that the FA-based defects (e.g. V FA , FA I and I FA ) have much lower formation energies. And anti-sites defects (e.g. FA I and I FA ) are located in the deep levels of bandgap, which can act as recombination centers and increase the V oc loss in PSCs. Furthermore, the defect formation energies in CsPbI 3 were calculated (figures 6(d) and (e)) by Liang et al [85] They found that CsPbI 3 has 12 point defects (the vacancies V Cs , V Pb , and V I , the interstitials Cs i , Pb i , and I i and anti-site occupations Cs Pb , Cs I , Pb Cs , Pb I , I Cs , and I Pb ). Under the Pb-poor condition, I i and V Pb possess the lowest formation energies among all donor and acceptor defects, respectively. While under the Pb-rich condition, I i and V I possess the lowest formation energies among all donor and acceptor defects, respectively.
In addition to theoretical calculation, several experimental techniques have been used to study the defects in perovskites [86][87][88]. Doherty et al used photoemission electron microscopy (PEEM) to image the trap distribution in perovskite films (figure 6(f)) [89]. The discrete, nanoscale trap clusters at the interfaces were observed. The regions with high photoluminescence (PL) efficiency showed little photo-excited hole trapping. And in regions with low PL efficiency, they see complex spatio-temporal dynamics with photo-excited holes being trapped at several discrete sites. Ni et al used drive-level capacitance profiling (DLCP) method to investigate the distributions of trap states in perovskite singlecrystalline and polycrystalline solar cells (figures 6(g)-(j)) [90]. Most of the traps are distributed on the top and bottom surfaces of the single crystal. The trap density of the singlecrystal was much lower than that of the polycrystalline films. Besides, the trap densities at the interfaces of the polycrystalline films were one to two orders of magnitude greater than that of the film interior. The results indicate that passivation of surface/interfacial defects may be the crucial issue to realizing the high performance and stability of PSCs.
Ions such as halogen [91,92], Li + [93] and Au/Ag [94] in PSCs can migrate under an external electric field, light, or heat mediated by vacancies and interstitial defects [95]. This usually leads to the failure of the perovskite device and cannot be solved by the encapsulation method. In figure 7(a), Domanski et al studied the migration of Au ion after heating the perovskite devices at different temperatures by TOF-SIMS (figure 7(b)) [94]. They showed that Au ion gradually diffused into the perovskite film and finally accumulated at the interface between ETL and perovskite upon increasing temperature. Kato et al detected the migration of I − and corrosion of the Ag electrode by using XRD and XPS (figure 7(c)) [91]. Under humid air condition, the migration of I − would be accelerated and moved through spiro-OMeTAD to the Ag electrode. The corrosion of Ag would occur by reacting with I − . Finally, the AgI would be formed.

Phase transition.
Phase instability of FA-based perovskites is the major obstacle hindering the solar cell longterm stability. The origin of phase instability is ascribed to the mismatched ionic radius [39], resulting in an inappropriate tolerance factor (∼1.0) and the distortion of crystal structure. The photo-active black phase is obtained at high temperat- [96,97]. And the black phase FAPbI 3 is metastable at room temperature, which turns to the yellow phase with the acceleration of the external condition (e.g. moisture). As a comparison, the ionic radius of MA + is suitable and the photo-active phase can maintain in MAPbI 3 perovskite. However, MAPbI 3 suffers from thermal instability which has been discussed in section 2.2.1.
Masi et al summarized the phase types of FAPbI 3 (figure 8) [96], including the photoactive black phases of α-(cubic), β-(orthorhombic), and γ-(tetragonal) and the photoinactive yellow phase of δ-(hexagonal). At room temperature, the δ-phase is the one with the lowest free energy of formation. When the temperature is increased, the photoactive δ-phase will convert into the black phase with the expansion of the unit cell volume and the increase of the Pb-I-Pb tilting angle (ϕ). Moreover, Chen et al used neutron diffraction and first-principles calculations to study the structure of FAPbI 3 perovskite [98]. They obtained the cubic phase at 390 K, a hexagonal phase at 220 K, and another hexagonal phase at 15 K to systematically study the function of FA + in the FAPbI 3 crystal structure and the importance of organic cations. They claimed that the entropy contribution to the Gibbs free energy caused by isotropic rotations of the FA + cation plays a crucial role in the cubic-to-hexagonal structural phase transition. Stabilizing the metastable perovskite (black phase) of FAPbI 3 is crucial for photovoltaic applications.

Phase separation.
Mixed halide perovskites are usually used to regulate the tolerance factor and improve the phase stability of FAPbI 3 . However, a new problem of phase separation appears in mixed halide perovskites, which becomes significant when increasing the Br − to I − ratios. Phase separation was first observed and reported by Hoke et al [99] They found that an additional PL peak forms at 1.68 eV on MAPb(Br x I 1−x ) 3 , the intensity of which grew under Reprinted with permission from [96]. Copyright (2020) American Chemical Society.
continuous illumination ( figure 9(a)). The position of this new peak is independent of halide composition (figure 9(b)). After continuous visible-light soaking of less than a minute, PL intensity from the new low-energy peak becomes more than an order of magnitude higher than the original peak (figure 9(c)). Furthermore, Li et al also observed the phase separation in all-inorganic CsPbIBr 2 films by using cathodoluminescence (CL) photomultiplier tube (PMT) mapping and secondary electron (SE) image [100]. The different colors between the grain boundaries and interiors indicated the formation of the I-rich phase at the grain boundaries. The I-rich phase would also segregate as clusters inside the films (figures 9(d) and (e)). The mechanism of the phase separation in CsPbIBr 2 was also explored. The mobile ions generated by the phase separation moved along grain boundaries as the path of ion migration.
Although the essential reason for phase separation is complex, it is widely accepted that the halogen-vacancy [80,81,101] and the excess charge-carrier [102] induced ionic migration in the perovskite films may be the main mechanism. Phase separation of perovskites happens either during the fabrication process or the aging period upon external stimuli (light [103,104], bias [105], etc). Bischak et al reported that the separation of halide in the perovskite films with light soaking is distinct from macroscopic phase separation. They observed the light-induced phase separation in MAPb(I 0.1 Br 0.9 ) 3 by CL imaging and multiscale simulations (figures 9(f) and (g)). The localized strain induced by the interaction between a photoexcited charge and lattice is sufficient to accelerate halide phase separation. And the low-bandgap, I-rich clusters are aggregated at the grain boundaries. Braly et al observed that the Br-rich composition of MAPb(I 0.6 Br 0.4 ) 3 and (MA 0.9 Cs 0.1 )Pb(I 0.6 Br 0.4 ) 3 experience rapid phase segregation upon 1-Sun equivalent current injection [105].
The pH value of the precursor solution affects the formation of I 2 impurity in perovskite films, which acts as the deeplevel defect and accelerates the degradation of perovskites under operational condition. As a consequence, regulating pH value of precursor solution is a rational method to stabilize perovskites [117]. Chen et al suppressed the formation of I 2 impurity by creation of an alkaline environment (figure 10(a)) [107]. The alkaline slowed down the crystallization kinetics of perovskite and retarded the formation of I 2 impurity ( figure 10(b)). By using a residual-free weak alkaline (FAAc) as an additive, the device showed a PCE of 20.87% and less degradation after storing in N 2 condition for 1500 h. Zhang et al fabricated FAPbI 3 films by using the pre-synthesized FAPbI 3 powder to reduce the defect density [41]. The defect density in the perovskite films was reduced,   Furthermore, organic additives could be used to stabilize the perovskite precursor solution. Qin et al found that the small molecule (ITIC-Th) could facilitate the incorporation of MA cation and suppress the formation of yellow phase FAPbI 3 (figures 10(g)-(i)) [48]. The device with ITIC-Th showed improved stability (figure 10(j)). Wang et al introduced the triethyl borate into the FAMA mixed precursor solution to restrain the deprotonation of MAI. The device showed better performance and reproducibility. Liu et al added the MA/EtOH into the precursor solution to suppress the formation of I 2 impurities and coordinate with Pb 2+ [44]. MA could adjust the perovskite colloidal size and stabilize the perovskite films. Li et al introduced the acetonitrile as the additive into the precursor solution to regulate the colloidal size and control the crystallization process. They obtained the device with a PCE of 19.7% [118].
The solvent is another crucial factor influencing the state of precursor solution and the quality of perovskite films. Many novel solvents and solvent mixture systems have been exploited. Chao et al introduced the room-temperature molten salt, methylammonium acetate (MAAc), as the solvent to prepare the MAPbI 3 precursor solution ( figure 11(a)) [114]. The hydrogen bonds between methylammonium, lead salts and MAAc caused complete solubility of the solute. Highquality perovskite films could be processed in ambient air.
The PSCs showed over 20% PCE and remained above 93% of original efficiency after storage in ambient air for more than 1000 h ( figure 11(b)). Noel et al used the low-viscosity, low boiling point solvent system (i.e. methylamine (MA) and ACN mixture) to prepare large-scale perovskite films (figure 11(c)) [115]. The uniform, pinhole-free perovskite films were fabricated ( figure 11(d)). Deng et al exploited a mixed solvent system containing 2-methoxyethanol (2-ME), ACN and dimethyl sulfoxide (DMSO) that could be applied for various perovskite compositions [116]. By tailoring solvent coordination capability, they obtained uniform perovskite films by blade-coating at an unprecedented speed of 99 mm s −1 . The device showed a PCE of 16.4% (63.7 cm 2 ) and over 1000 h operational stability. Yoo et al employed 2-methoxyethanol (2-ME) as the solvent of precursor solution (figure 11(e)) [63]. Highly uniform and pinhole-free perovskite films could be achieved by adding n-cyclohexyl-2-pyrrolidone (CHP) into the 2-ME solution. They obtained the champion laser-patterned perovskite mini-module (figure 11(f)) with a PCE of 20.4% (31 cm 2 ). What's more, the mini-module was stable under ambient conditions for 50 d.

Composition regulation.
The metal cations were introduced into the perovskite films as dopants with the aim to improve either efficiency or stability of PSCs. The performance and stability of PSCs with cations regulation were summarized in table 2. Among metal cation dopants, alkali metal cations have shown to improve the crystal structure stability and the crystallinity of the perovskite films [30,58,[119][120][121][122][123][124]. In 2016, Saliba et al reported that the small and  oxidation-stable Rb + can be embedded into a 'cation cascade' to form multi-cation perovskite film with excellent optoelectronic properties ( figure 12(a)) [125]. The device showed 21.8% PCE and retained 95% of its initial performance after soaking under full sunlight at 85 • C for 500 h. A similar phenomenon was observed when using Cs + as the dopant in perovskite films [126]. In 2018, they incorporated both the Rb + and Cs + into FAPbI 3 films to realize highly crystalline formamidinium-based perovskites without any Br − or MA + ( figure 12(b)) [127]. They obtained PSCs with an efficiency of 20.35%. Besides, the polymer-modified device maintained over 98% of the initial PCE after 1000 h of continuous maximum power point (MPP) tracking in a nitrogen atmosphere. The K + was incorporated to reduce the J-V hysteresis and improve stability of the perovskite device [119]. Abdi-Jalebi et al found that the I − in KI could compensate for the halide vacancies and the K + could combine with the halides at the grain boundaries and surfaces, thereby inhibiting halide migration and suppressing the additional non-radiative decay arising from interstitial halides (figure 12(c)) [128]. Bu et al reported a universal potassium interfacial passivation strategy to improve the interfacial stability ( figure 12(d)) [123]. The potassium passivated device showed excellent light stability and long-term storage stability.  [131], etc have also been doped into the perovskite films to improve their thermal and illumination stability. Saidaminov et al found that the CdCl 2 could suppress the atomic vacancies in perovskite films (figure 12(e)) [71]. Doping with CdCl 2 improved the stability of PSCs by an order of magnitude. Wang et al reported that the Eu ion pair Eu 3+ -Eu 2+ could act as the 'redox shuttle' that selectively oxidized Pb 0 and reduced I 0 defects simultaneously via a cyclical transition process (figure 12(f)) [129]. They obtained the device with a certified PCE of 20.52%. In addition, the Eu-containing devices maintained 90% of the original PCE even after 8000 h of storage and 91% of the original stable PCE after MPP tracking for 500 h.
Anions were incorporated into the perovskite films as the Lewis-acid to passivate the Pb interstitial defects, undercoordinated Pb 2+ and halide-vacancy defects. Generally, halide ions with a small radius (e.g. F − and Cl − ) can be used to improve the crystallinity, passivate the vacancy defects and enhance the stability of the perovskite films [24,132]. On the other hand, halide ions with a suitable radius (e.g. Br − ) can replace I − in the [PbI 6 ] − octahedron to adjust the bandgap and improve the phase stability [133,134]. The performance and stability of PSCs with anions doping were summarized in table 3. Li et al employed fluoride to simultaneously passivate the anion and cation vacancy defects ( figure 13(a)) [132]. They found that the extremely high electronegative of fluoride could enhance the hydrogen bond and ionic bond in the perovskite structure. With the incorporation of fluoride, the device demonstrated a high PCE of 21.46% and improved stability under stresses such as illumination, heat and humidity ( figure 13(b)). Min et al reported a new interlayer (i.e. FASnCl x ) between the Cl-doped SnO 2 and Clcontaining perovskite films [24]. The interfacial defects and charge extraction were improved because of the atomically coherent features of the interlayer. With the interfacial passivation of Cl − , they fabricated a device with a certified PCE of 25.5%. Besides, the unencapsulated device maintained ∼90% of the initial PCE after light exposure for 500 h. Graphene has been widely used to impede ion migration and perovskite decomposition [135,136]. Wang et al modified the surface of the perovskite with the Pb(SCN) 2 and chlorinated graphene oxide to construct the Pb-rich surface and the strong Pb-Cl and Pb-O bonds [137]. The loss of decomposed components could be extremely impeded with the coverage of chlorinated graphene oxide and the operation stability could be enhanced. The device maintained 90% of its initial PCE after operation at MPP under AM1.5G solar light at 60 • C for 1000 h.
Moreover, the non-halogen anions have been used to suppress the anion-vacancy defects at the grain boundaries and on the surface [138]. Jeong et al introduced pseudo-halide anion formate (HCOO − ) into the perovskite precursor solution, which is small enough to fit into the perovskite structure and fill the I-vacancy [59]. They obtained the perovskite films with grain sizes of up to 2 µm (figure 13(c)) and PSCs with a PCE of 25.6% (certified 25.2%). The shelf-life, heat, and long-term operational stability were extremely improved ( figure 13(d)). The sulfate or phosphate ions have also been used to passivate the defects and stabilize the perovskite  [139]. The wide-bandgap lead oxysalt could reduce the trap density on the perovskite surface by passivating the undercoordinated surface lead. As a consequence, the device maintained 96.8% of its initial efficiency after operation at MPP under simulated air mass (AM) 1.5G irradiation for 1200 h at 65 • C ( figure 13(f)).
Quasi-2D structured perovskites have presented better stability than 3D perovskites due to the higher hydrophobicity of the large organic cations and the higher formation energy [140]. Ruddlesden-Popper (RP), Dion-Jacobson (DJ) and the alternating cations in the interlayer space (ACI) phase are the most common 2D perovskites which are oriented along the (100)-plane [141,142]. Though the stability of the quasi-2D perovskite is much more attractive [143], the performance of the 2D PSCs is unsatisfactory compared with the 3D PSCs. For the RP structured 2D perovskites, the van der waals gap between the adjacent inorganic [PbI 6 ] 4− sheets caused the formation of a large interlayer space, influencing the carrier transport across the inorganic layers. Moreover, the disorder of the quantum wells (QW) distribution and film growth orientation also inhibited the carrier transport [144]. For the DJ structured 2D perovskites, van der waals gap is avoided. However, the QW distribution and the crystallization of the DJ perovskite would affect the performance of the PSCs [145,146]. The ACI perovskites are a new type of 2D halide perovskites that featured two different alternating cations in the interlayer space. Recently studies showed that the quality of the ACI perovskite films and the crystallization kinetics may determine the performance of the PSCs [142,147]. The performance and stability of PSCs based on these 2D organic cations has been summarized in table 4.  A large number of research works have focused on the RP structured PSCs [148][149][150][151][152][153]. Ren et al demonstrated that a sulfur-sulfur interaction was presented for a new bulky alkylammonium, 2-(methylthio)ethylamine hydrochloride (MTEACl) ( figure 14(a)) [154]. The interaction between sulfur atoms in two MTEA molecules enabled the (MTEA) 2 (MA) 4 Pb 5 I 16 (n = 5) perovskite framework with enhanced charge transport and stability. The PSC obtained 18.06% PCE, better moisture tolerance (1512 h under 70% humidity conditions), higher thermal stability (375 h at 85 • C) and operational stability under continuous illumination (85% of the initial efficiency retained over 1000 h). Liang et al fabricated phase-pure QWs by introducing the molten salt spacer n-butylamine acetate ( figure 14(b)) [155]. The highquality phase-pure QW perovskite films in the 2D RP structure could be obtained by the strong ionic coordination between n-butylamine acetate and the perovskite framework. They obtained the PSCs with a PCE of 16.25% (n = 4). The device was stable under 65 ± 10% humidity for 4680 h, heating at 85 • C for 558 h) and light illumination for 1100 h. Dong et al introduced the 2-thiopheneformamidinium (ThFA) as the organic spacer to prepare the 2D (ThFA) 2 (MA) n−1 Pb n I 3n+1 (figure 14(c)) [156]. The 2D perovskite showed preferential vertical growth orientations, high charge carrier mobilities, and reduced trap density. The 2D RP device exhibited a high PCE of 16.72% with a low n-value of ∼3. The device also presented improved stability with less than 1% degradation after storing in N 2 for 3000 h. DJ structured perovskites have been developed to avoid the van der waals gap and enhance the carrier transport between the inorganic sheets [145,[157][158][159]. Zhang et al developed a DJ structured perovskite with the composition of (PDMA)(MA) n−1 Pb n I 3n+1 (n = 4, PDMA refers to 1,4phenylenedimethanammonium) ( figure 14(d)) [141]. They found that the uniform thickness distribution of QWs could be obtained by the hot-casting or antisolvent processes. Using the hot-casting method, they prepared a DJ device with PCE reaching 15.81%. The thermal and humidity stabilities of the 2D perovskites were extremely enhanced compared with the 3D MAPbI 3 perovskite. Lv et al introduced a multifluorinated aromatic spacer, namely 4F-PhDMA, into the 2D DJ perovskite (figure 14(e)) [144]. The DJ perovskite with 4F-PhDMA spacer exhibited higher dissociation energy compared with the PhDMA because of the multiple noncovalent interactions such as NH···I and CH···F hydrogenbonding and F···I electrostatic interactions. By incorporating the 4F-PhDMA organic spacer, they obtained the device (n = 4) with a PCE of 16.62%, advanced storage stability (>93% after storing in N 2 for 1839 h), operational stability (∼94% under continuous-light illumination for over 700 h) and heating stability (>94% under 80 • C after 350 h). Di et al synthesized a thiophene-based bulky dication spacer, namely 2,5-thiophenedimethylammonium (ThDMA), to fabricate the high-quality 2D DJ perovskite (ThDMA)MA n−1 Pb n I 3n+1 (nominal n = 5) (figure 14(f)) [160]. With the strong coordination molecule DMSO, The crystal growth and orientation of the 2D DJ perovskites could be enhanced, enabled by strong coordination between ThDMA and DMSO. A high PCE of 15.75% was achieved and the DJ device exhibited much better storage stability, light soaking stability and thermal stability than the 3D counterparts.
ACI structured 2D perovskite was firstly reported by Soe et al in 2017 ((GA)MA n Pb n I 3n+1 (GA = guanidinium, MA = methylam-monium)) [161], and has become one of the hotspots in 2D perovskites [142,147,162]. Zhang et al carried out in situ studies on the solidification processes of ACI 2D perovskite (n = 3) by using in situ grazing-incidence x-ray scattering (GIWAXS) [142]. They found that the intermediate phases, e.g. 2D GA 2 PbI 4 perovskite, provided a scaffold for the growth of the ACI perovskites during thermal annealing. Yang et al tailored the crystallization process of ACI perovskite (i.e. (GA)MA n Pb n I 3n+1 ) via solvent engineering to achieve preferential QW distribution and improve the quality of perovskite films. The PSC obtained a high PCE of 19.18% and high environmental stability [147].

Strain engineering. Tensile strain in perovskite films
is an important source of instability. Many works have been reported for regulating the strain in perovskite devices, such as interfacial modification [163], doping [60,71] and employment of suitable transport materials [73]. Dou et al introduced an ultrathin Eu-MOF layer between the electron transport layer and perovskite layer. The tensile strain in perovskite film was successfully converted into the compressive strain (figures 15(a) and (b)) [164].  figure 15(c)) [60]. They found that adding a 0.03 mol fraction of both MDA and Cs cations could lower the lattice strain. The PSC reached a high PCE of 24.4% and maintained over 80% of the initial PCE after heating at 85 • C for 1300 h. Zhang et al regulated the strain by a crosslinking-enabled strainregulating crystallization (CSRC) method ( figure 15(d)) [165]. A suitable concentration of the trimethylolpropane triacrylate (TMTA) was used to convert a tensile strain into strain-free perovskite film. The device with TMTA retained 80% of the initial PCE after light soaking for 1248 h ( figure 15(e)). Wang et al used OAI as the A-site cation to release the interfacial stress ( figure 15(f)) [166]. Soft structural subunits were realized and a 'bone-joint' configuration was constructed at the interface between the absorber and the carrier transport layer. The treatment of OAI led to improved humidity and thermal stability ( figure 15(g)).

Defects passivation.
Polycrystalline perovskite films exhibit a variety of defects, i.e. positively charged cationic defects (under-coordinated Pb 2+ , and halide vacancies), negatively charged anionic defects (cation vacancies, Pb-I antisites and halide-excess) and metallic lead (Pb 0 ) defects, etc. These defects are ion migration channels and cause instability of the PSCs. In this section, we summarized the recent defect passivation methods for improving the stability of the PSCs.
Lewis acid is an electron-pair acceptor which can accept a foreign electron pair [167]. Besides the anions, the organic molecule with the Lewis-acid functional group, such as TPFPB [168], PCBB-3N-3I [169], TPFP [170], iodineterminated self-assembled monolayer (I-SAM) [171], I-PEA [172], TMOS [173], PFTS [174], etc, could also be introduced to passivate the halide-vacancy and Pb-I antisite defects in perovskite and enhance the stability of the PSCs. The performance of perovskite device based on the Lewis acid-base passivation is summarized in table 5. Fu et al introduced a halogenhalogen bond at the grain boundaries of perovskite to suppress the ion migration and the phase separation ( figure 16(a)) [172]. They found that the halogen atom with a positively charged hole acts as an electron acceptor (Lewis acid) and forms strong halogen-halogen bonds with the electron-rich halide anions (Lewis base). The binding energy of the halogen-halogen bond is higher than that of the hydrogen bond ( figure 16(b)). The halogen-halogen bond-containing CsMAFAPb(I x Br 1−x ) 3 perovskite films enabled the encapsulated device to retain 90% of initial PCE after MPP tracking for 500 h. Dai et al found that the I-SAM could be used as the Lewis acid to form the electrostatic bonds with the perovskite films (figure 16(e)) [171]. Treatment of the buried interface with the I-SAM significantly suppressed the point defects in perovskite, leading to improved solar cell performance. The T 80 of the device reached 4000 h under 1-sun illumination with MPP tracking, benefiting from reduced ion migration.
Lewis base is an electron-pair donor. It may provide the self-electron pair and coordinate with the Lewis acid type of defects (i.e. interstitial Pb 2+ and under-coordinated Pb 2+ defects) [167]. Organic molecules with the Lewis base functional groups, such as O-donor [175][176][177][178][179][180][181][182], S-donor [183][184][185][186][187], N-donor [188][189][190], have been widely used as Lewis acid defect passivators. Wang et al reported that the C=O and N-H groups in the theophylline, caffeine and theobromine could act as the Lewis base to passivate the antisite Pb defect to maximize surface-defect binding ( figure 16(c)) [176]. They obtained the stabilized PCE of 22.6% with theophylline modified perovskite (FAPbI 3 ) x (MAPbBr 3 ) 1−x . The encapsulated device maintained over 90% of its initial efficiency under continuous light at the open-circuit condition for 500 h. Zhao et al found that dimethyl itaconate (DI) with C=C and C=O functional groups could interact with PbI 2 while the polymerization triggered by the annealing process ( figure 16(d)) [177]. The polymerization would be adhered to the grain boundaries and passivate the under-coordinated Pb 2+ defects. The FA 1−x MA x PbI 3 PSCs obtained 23% PCE and prolonged lifetimes when 85.7% and 91.8% of the initial PCEs remained after 504 h continuous illumination and 2208 h shelf storage. Zhang et al introduced the organic molecule with the S, N functional group to combine with Pb 2+ in the perovskite film (figure 16(f)) [183]. The PCE of the 2-MPpassivated device achieved 20.28% efficiency. The unencapsulated device retained 93% of the initial efficiency under a RH of 60%-70% for 60 d.
Furthermore, the zwitterion with the Lewis acid and base functional groups showed an excellent bilateral effect by simultaneously passivating negatively and positively charged defects [167,[191][192][193][194]. Cao et al reported the employment of a star-shaped polymer at the perovskite interface to improve charge transport and inhibit ion migration [192]. The polymer with the Lewis acid functional group -CF 3 and the Lewis base functional group C=O formed the strong hydrogen bond of F…H-N between the polymer and FA + or MA +, and the coordination between C=O and Pb 2+ . The crystallization process of perovskite film was well controlled and the films were prepared with lower trap density and higher carrier mobility. They obtained the device with a PCE of 22.1% and a FF of 0.862. The device exhibited excellent environmental stability, long-term operational stability and thermal stability. Choi et al employed the zwitterionic, 3-(1-pyridinio)-1-propanesulfonate, to modify the interface between SnO 2 and perovskite ( figure 16(g)) [193]. The zwitterion at the interface could passivate the Pb-I antisite defects and enhance the electron transport ability of SnO 2 . Finally, the device showed a PCE of 21.43% and excellent stability under 85 • C, 85% RH.
Besides the Lewis-acid and base salts, organic ammonium salts have been widely used to simultaneously passivate the cation and anion defects by doping or interfacial modification. The role of the organic ammonium salts on the perovskite films mainly consists of the following aspects: (I) short-chain organic ammonium salts could passivate the defects, inhibit the ion migration and reduce leakage current loss through forming the wide-bandgap low-dimensional perovskite [195][196][197][198][199][200][201][202][203][204]; (II) long-chain organic ammonium salts could improve energy level structure, optimize the growth orientation of the perovskite films and enhance the humidity and   [205][206][207][208][209]. The performance and stability of PSCs by organic-ammonium salts passivation has been summarized in table 6. Short-chain organic ammonium salts were employed to passivate the surface defects and form the 2D wide-bandgap perovskite layer on the surface of 3D perovskite layer. Jiang et al found that the organic molecule phenethylammonium iodide (PEAI) could serve as a much more effective passivator than the traditional PEA 2 PbI 4 for the FA-MA mixed perovskite films through the non-annealing process (figure 17(a)) [210]. The device obtained a certificated efficiency of 23.32% and a high V oc of 1.18 V. Besides, the device was stable under 85 • C for 500 h and continuous light soaking at the MPP (25 • C, 100 mW cm −2 ) for over 40 h. Proppe et al introduced the ligand 4-vinylbenzylammonium to form 2D perovskite quantum wells (PQWs) on the 3D perovskite layer ( figure 17(b)) [201]. Especially, the vinyl group of 4vinylbenzylammonium could be activated using 254 nm UV light to form new covalent bonds in the 2D PQWs. Based on the UV-cross-linked 2D/3D structure, they obtained the champion PCE of 20.4%. The device retained 90% of its initial efficiency after aging in dark for 2300 h and 75% after operating for 16 h. Jang et al employed a new method to prepare the 2D/3D heterojunction (figure 17(c)) [202]. They fabricated the perovskite films by growing a highly crystalline 2D (C 4 H 9 NH 3 ) 2 PbI 4 (n = 1) film on top of the 3D film without the quasi-2D phase using a solvent-free solidphase in-plane growth (SIG) method. The perovskite films demonstrated a prolonged carrier lifetime, well-passivated surface defects and enhanced build-in potential. They obtained the device with a high PCE of 24.35%. The encapsulated device retained 94% of its initial efficiency after 1056 h under the 85 • C/85% RH condition and 98% after 1620 h under full-sun illumination.
Long-chain organic ammonium salts were also used to passivate the defects and enhance the device stability, especially the humidity stability. Generally, the ability of electron blocking and humidity resistance characteristics is different for the organic ammonium salts with different chain lengths. Jung et al proposed a new device architecture for the high-performance P3HT-based PSCs ( figure 17(d)) [205]. The wide-bandgap perovskite layer could be formed with the reaction of n-hexyl trimethyl ammonium bromide (HTAB). The long alkyl chain (C 6 H 13 − ) in HTAB could form favorable van der Waals interactions between the P3HT and the perovskite layer and the functionalized moiety (N + (CH 3 ) 3 − ) in HTAB molecule could passivate the charge traps on the perovskite surface. As a result, they obtained a certified PCE of 22.7%. The device exhibited good stability at 85% RH without encapsulation, and excellent long-term operational stability by maintaining 95% of its initial efficiency under 1-sun illumination at room temperature for 1370 h with encapsulation. Reproduced from [193] with permission from the Royal Society of Chemistry.
Zheng et al incorporated a trace amount of surface-anchoring alkylamine ligands (AALs) with different chain lengths into the perovskite precursor solution to passivate the defects and optimize the crystal orientation (figure 17(e)) [207]. They found that the perovskite films with the low concentration of surface-ALLs doping exhibited a prominent (100) orientation and the alkylamine ligands could act as the AALs to optimize the carrier transfer (figure 17(f)). They obtained the device with 23% PCE and with operational stability of over 1000 h. Zheng et al showed that the quaternary ammonium halides could passivate the ionic defects more effectively compared with PCBM in various perovskite components (figure 17(g)) [211]. This not only reduced the V oc loss to 0.39 V but also boosted the PCE to 20.59 ± 0.45%. The device modified with choline chloride was stable for over 35 d in air and maintain 86% of the initial PCE under 1 sun continuous illumination for 26 h.
Li et al developed a liquid medium annealing (LMA) method to create a robust chemical environment-anisole for the crystal growth during the annealing process (figure 18(f)) [216]. They successfully prepared the perovskite films with high crystallinity, fewer defects and desired stoichiometry. The PSCs showed a PCE of 24.04% and 23.15% over areas    [202], with permission from Springer Nature. (d) Left, the structure of an n-i-p perovskite solar cell based on a double-layered halide architecture (DHA) using P3HT as the hole-transport material. Right, schematic structure of the interface between the wide-bandgap halide (WBH) and P3HT. Reproduced from [205], with permission from Springer Nature. (e) Illustration of the influence of the short-chain surface-anchoring alkylamine ligands (AALs) and long-chain AALs on the crystallization of the perovskite films. (f) Illustration of long-chain AALs assembled on the perovskite film surface and blocking the holes at the perovskite and C 60 interfaces. Reproduced from [207], with permission from Springer Nature. (g) Device structure and passivation mechanism by quaternary ammonium halides. Reproduced from [211], with permission from Springer Nature.
of 0.08 cm 2 and 1 cm 2 , respectively. The operational stability of the device was also extremely enhanced. Huang's group developed the adhesive tape treatment method (figure 18(g)) [217] and the polish method [218] to remove the defective surface layers. As shown in figure 18(h), the right tape-treated part remained black after 8 h of light soaking, while the left part without tape-treatment already decomposed into yellow phases in less than 4 h. The tape-treated device retained 97.1% of its initial PCE after operation near MPP under 1-sun illumination for 1440 h at 65 • C. The right half of the film was treated with adhesive tape. Reprinted from [217], Copyright (2020), with permission from Elsevier.

Phase stabilization.
FAPbI 3 is the most popular component for high-efficiency PSCs because of its suitable bandgap. However, the phase instability issue due to the ionic radius mismatch would inhibit the development of stable FAPbI 3 PSCs. Many methods, such as additive engineering, interfacial engineering, solvent engineering, etc, have been reported to stabilize the black-phase FAPbI 3 perovskite films [60,61,165,[219][220][221][222][223][224]. Kim et al doped the methylammonium chloride (MACl) into the FAPbI 3 precursor solution to stabilize the α-phase FAPbI 3 ( figure 19(a)) [225]. They found that the MA + and Cl − in MACl could be incorporated into the lattice of FAPbI 3 to reduce the formation energy α-phase FAPbI 3 and enhance the crystallinity of FAPbI 3 ( figure 19(b)). The optimized PSCs achieved a certified efficiency of 23.48% and the ideal stability. Xie et al used the MACl to modify the surface of FAPbI 3 films and assist the vertical recrystallization [226]. The FA-based device with the low MA content showed 20.65% PCE and 500 h thermal stability. Lu et al reported a methylammonium thiocyanate (MASCN) vapor treatment method to stabilize the α-phase FAPbI 3 [227]. The phase transition temperature was decreased with the treatment of MASCN ( figure 19(c)). Yoo et al incorporated the MAPbBr 3 and MACl into the precursor solution to stabilize the FAPbI 3 ( figure 19(d)) [134]. The MAPbBr 3 additive not only enhanced the crystallinity but also increased the effective mobility of FAPbI 3 . Based on the CBD processed SnO 2 and the passivation strategy, they obtained the 25.2% certified PCE (figure 19(e)) and excellent storage and illumination stability. Park et al used the isopropylammonium cations (iPAmH + ) to stabilize the FAPbI 3 ( figure 19(f)) [228]. They found that iPAmH + appeared by the chemical reaction between isopropyl alcohol (IPA) and MACl. The device with iPAmH + exhibited a PCE of 23.9% and better long-term operation stability.
Ionic liquid was widely used to enhance the crystallinity and stabilize the FAPbI 3 . Hui et al used a novel solvent (i.e. methylamine formate (MAFa) ionic liquid) to fabricate the stable α-FAPbI 3 ( figure 19(g)) [229]. The MAFa exhibited strong interactions with PbI 2 through C=O…Pb and N-H…I, which promoted the vertical growth of the FAPbI 3 crystals. The device showed the champion PCE of 24.1% and retained 80% of its initial PCE at 85 • C for 500 h. Akin et al employed the 1-hexyl-3-methylimidazolium iodide (HMII) ionic liquid as the additive to solve the phase instability of FAPbI 3 ( figure 19(h)) [230]. They found that the  [227]. Reprinted with permission from AAAS. (d) X-ray diffraction of perovskite thin films with four different amounts of MAPbBr 3 . (e) J-V curves of the champion device measured at Newport, showing both the conventional J-V sweep and the certified quasi-steady-state measurements. Reproduced from [134], with permission from Springer Nature. (f) Simulated formation energies of α-FAPbI 3 and δ-FAPbI 3 crystals in the bulk and thin-film phases. Reproduced from [228], with permission from Springer Nature. (g) Images of PbI 2 @MAFa and PbI 2 @DMF: DMSO solutions and schematic diagram of interactions in the solutions. From [229]. Reprinted with permission from AAAS. (h) Schematic illustrations of the device structure, molecular structure of the 1-hexyl-3-methylimidazolium (HMI + ) and the effects of HMI + in FAPbI 3 active layer. [ [225], Copyright (2019), with permission from Elsevier. Reproduced from [134], with permission from Springer Nature.
HMII could increase the grain size and reduce the activation energy of the grain-boundary migration. The FAPbI 3 device with the HMII additive achieved a PCE of 20.6% and maintained 80% and 95% of its initial PCE at 60 ± 10% RH and 65 • C, respectively.

Conclusion and outlook
Efficiency, cost and lifetime are the solar cell performance golden triangle [231]. Until now, the certified record PCE of PSCs has reached 25.5%. The energy payback time of PSCs is estimated to be 2-3 months [232]. The lifetime becomes the key obstacle hindering technology commercialization. In this review, we reviewed the origin of instability of the solution-processed PSCs from the perspectives of the precursor solutions (i.e. solute and solvent) and the perovskite films (i.e. composition, strain, defect and phase). In addition, we summarized the recent strategies for improving stability of the perovskite films and solar cells, including perovskite precursor solution advancement, perovskite composition regulation, strain engineering, defect passivation and phase stabilization. The perspective for the stable PSCs was depicted in figure 20. Furthermore, we proposed the following strategies that may hold the key to further enhancing the stability of perovskite films and devices.
(a) Exploration of novel reductants in perovskite precursors to inhibit the oxidization of I − . One of the key degradation pathways of the perovskite precursor solution is the oxidization of I − . The formation of I 2 not only affects the reproductivity of the high-quality perovskite films but also deteriorates the long-term stability of the perovskite devices. The development of novel reductants in the precursors which effectively inhibit the oxidization of I − will prolong the durability and guarantee period of the perovskite precursor solution. This is a prerequisite when solution-processed perovskites are to be considered at an industrial-relevant manufacturing scale. (b) Regulation of the crystallization kinetics for the preparation of high-crystallinity (or single-crystalline-like) perovskite films with low defect densities and long carrier lifetime. Crystallinity and defect in perovskite semiconductors play a vital role in solar cell performance and stability [32]. The high density of defects and short carrier lifetimes usually exist in the perovskite films with low crystallinity, resulting in large electric loss and severe ion migration. The formation of defects (i.e. type, density and location) is determined by the crystallization kinetics. From this viewpoint, fundamental understanding of crystallization kinetics and defect chemistry allows the development of advanced perovskite precursor solutions and the fabrication of high-crystallinity perovskite films. (c) Employment of suitable A-site cations for enabling quasi-2D perovskites with optimized crystal growth orientation and enhanced carrier transport. Quasi-2D structured perovskites have presented higher stability than 3D perovskites but unsatisfactory carrier transport characteristics. Previous studies showed that A-site cations affect the growth orientations of perovskites, which determine carrier transport and stability of the device. Understanding the effect of structures of A-site cations (e.g. length of the hydrophobicity chains and the functional groups) on the perovskite growth orientation, carrier and ion transport kinetics lay the foundation for high-efficiency and stable quasi-2D PSCs. (d) Advancement of other functional layers (i.e. transport layers and electrodes) which demonstrate high intrinsic robustness and are chemically inert with the perovskite layer. Transport layers and electrodes are important components in PSCs and influence the long-term stability of PSCs. The mostly investigated hole transport materials , e.g. spiro-OMeTAD, are unstable under heating or moisture [233]. The electron transport materials, e.g. TiO 2 , promote the decomposition of perovskite films under light illumination [234]. The Ag and Au electrodes react with the mobile halide ions, causing irreversible device degradation [91]. Therefore, designing dopant-free charge transport layers (e.g. inorganic semiconductors and conductive polymers) with high-temperature and humidity resistance and tailoring the interfaces between perovskite and charge transport layers may promote the stability of PSCs [235,236]. Furthermore, the employment of stable electrodes (i.e. transparent conductive oxides and carbon) or the construction of double-layer metal electrodes (i.e. Bi/Au and ITO/Au) avoid the detrimental metal halide reactions. (e) Development of advanced encapsulation methods to avoid the ingress of oxygen and moisture, and outgassing of perovskite degradation gaseous species. External encapsulations are key components of commercial photovoltaics such as silicon, copper indium gallium selenide and CdTe solar cells. The degradation pathways of PSCs induced by external stimuli (moisture, heat and oxygen) have been well-documented, and can be mitigated by reliable external encapsulation methods [237][238][239]. Several encapsulants (e.g. epoxy [240] and polyisobutylene [237]) and encapsulation strategies (e.g. edge sealing and blanket encapsulation [53]) have shown promising results under independent accelerated stress conditions including continuous light-soaking, damp heat and thermal cycling. In future, exploration of advanced encapsulation methods of PSCs is to be investigated, aiming to pass harsher combination stresses such as simultaneous light, temperature and RH [33].