The stability of inorganic perovskite solar cells: from materials to devices

Inorganic halide perovskite solar cells (IHPSCs) have become one of the most promising research hotspots due to to the excellent light and thermal stabilities of inorganic halide perovskites (IHPs). Despite rapid progress in cell performance in very recent years, the phase instability of IHPs easily occurs, which will remarkably influence the cell efficiency and stability. Much effort has been devoted to solving this issue. In this review, we focus on representative progress in the stability from IHPs to IHPSCs, including (i) a brief introduction of inorganic perovskite materials and devices, (ii) some new additives and fabrication methods, (iii) thermal and light stabilities, (iv) tailoring phase stability, (v) optimization of the stability of inorganic perovskite solar cells and (vi) interfacial engineering for stability enhancement. Finally, perspectives will be given regarding future work on highly efficient and stable IHPSCs. This review aims to provide a thorough understanding of the key influential factors on the stability of materials to highly efficient and stable IHPSCs.


Introduction
In recent years, perovskite solar cells (PSCs) based on organic-inorganic hybrid lead halide light absorbers have * 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. become one of the most focused research fields in the photovoltaic field due to their outstanding photoelectric conversion properties [1][2][3][4]. Since the first PSC was reported by Miyasaka et al in 2009, the power conversion efficiency (PCE) of PSCs has already climbed to 25.7%, comparable with that of the state-of-the-art crystalline silicon solar cells [5][6][7][8][9][10][11][12][13]. To date, the PSCs with higher PCEs over 22% have all been based on organic-inorganic hybrid perovskite materials, which exhibit a 3D crystal structure with a general chemical formula of ABX 3 , where A is a monovalent cation, such as methylammonium (CH 3 NH 3 + , MA + ) and Crystal structure of a typical black phase ABX 3 perovskite. Reproduced from [14] with permission from Springer Nature.
formamidinium (HC(CH 2 ) 2 + , FA + ), B is a divalent cation, such as Pb 2+ and Sn 2+ , and X is a halide ion, such as I − , Br − , and Cl − , as shown in figure 1 [14]. However, the PSCs based on organic-inorganic hybrid perovskite face a series of stability issues. For example, compositional and structural degradation, thermal and chemical instabilities derived from phase transformations (i.e. moisture or oxygen) and UV light instability [15][16][17][18]. Extensive attention has been focussed on improving stabilities of PSCs to promote its commercial application. For example, introducing additives to control the perovskite crystallinity and to reduce defects and grain boundaries, developing interfacial layers to passivate interface defects, or designing new materials from the perovskite itself, hole transporting materials to carbon electrodes [13,[19][20][21][22][23][24][25][26][27][28][29][30][31]. In addition to organic ammonium cations, inorganic cations, such as Cs + , can also be used to construct a 3D perovskite with Pb 2+ and halides, thus forming allinorganic perovskite materials. For example, CsPbI 3 , CsPbBr 3 and their alloying compounds [32][33][34]. These materials also exhibit a high visible-light absorption coefficient, thus being able to be used as a light absorber for the cell [33]. Due to a much lower volatility and reduced rotation freedom of the Asite inorganic cation, thermal and light illumination stabilities of all-inorganic perovskite materials have been significantly improved [35][36][37][38][39]. Thus, it is generally considered that allinorganic PSCs have provided more alternative approaches for overcoming the stability issue of PSCs.
The appearance of inorganic perovskite materials could date to 1893. However, it was not until 2015 that CsPbI 3 and CsPbBr 3 were first assembled into photovoltaic devices (so-called inorganic halide perovskite solar cells (IHPSCs)) [32,40]. Compared to hybrid perovskite cells, these IHPSCs exhibit superior operational stability under continuous light illumination and electric field [38,41,42]. By in-depth study of the composition engineering, crystallization control and interface engineering of IHPSCs, its PCEs have already climbed to 21% [39,[41][42][43][44][45][46][47]. These achievements have been mainly dependent on the development of inorganic perovskite materials and further understanding device operation principles. In addition to CsPbX 3 perovskites, other all-inorganic perovskites have also been developed [48,49]. In recent years, the phase stability issue of the all-inorganic perovskite has been of great concern, and numerous efforts have been made to improve this stability [37].
In this review, we focus on the important progress related to stabilities from inorganic halide perovskites (IHPs) to IHPSCs. Meanwhile, more attention is also paid to phase instability and accurate control. Finally, perspectives for future development of highly efficient and stable IHPSCs are presented.

Inorganic perovskite materials and devices
In this section, the crystal structure and properties of IHPs are presented first. We then provide a brief introduction to the architectures and fabrication methods of IHPSCs.

Structure and properties of IHPs
Cs-based IHPs are described by the chemical formula CsBX 3 (B = Pb 2+ , Mn 2+ , Sn 2+ , etc, X = Cl − , Br − , I − ), where Cs + occupies a corner of a unit cell with 12-fold coordination, B is a divalent metal cation sitting in a body-centered position with six-fold coordination, and X is a halogen in a face-centered position, as depicted in figure 1 [14]. Generally, the crystal structure of the CsBX 3 is determined by two important empirical parameters, Goldschmidt's tolerance factor t and the octahedral factor µ, which are defined as follows [50,51]: where r represents the ionic radius of the corresponding cations and anions. To the phase-stable halide perovskites, the proper range of the tolerance factor is 0.8 < t < 1. Furthermore, µ in the range from 0.4-0.9 ensures the formation of stable [BX 6 ] 4− octahedra. Although the ionic radius of Cs + is smaller than those of organic MA + and FA + cations, their overall phase stability is not far from the above-mentioned range, as shown in figure 2(a). Due to these structural limitations (simultaneously satisfying the t and µ conditions), when the A site is occupied by Cs + , the combination numbers of B and X ions are very limited to 3D perovskite, as shown in figure 2(b). Under an ideal situation, if the void size determined by [BX 6 ] 4− octahedra matches well with the Cs + size, the ideal cubic perovskite phase with t = 1 is obtained. Unfortunately, the Cs + size is smaller than the void size, leading to t < 1. For example, the reported tolerance factor of CsPbI 3 is 0.807, which has to undergo a higher octahedral rotation distortion in comparison to CsPbBr 3 with a higher t value (t = 0.824). In other words, the [PbX 6 ] 4− octahedron will rotate and tilt to reduce the extra space around the Cs + cation. In addition, taking   [52], reprinted with permission from AAAS. (b) Crystal structure and relative phase transition process of different phases. From [52], reprinted with permission from AAAS. CsPbI 3 as an example, a thermodynamically stable yellow δphase (non-perovskite) at room temperature will transform into an optically active black perovskite phase through a reversible high-temperature phase: α (cubic), β (tetragonal) and γ (orthorhombic). This process involves the rotation of the [PbI 6 ] 4− octahedron. Thermal phase relations of CsPbI 3 are depicted in figure 3 [52]. The term 'black' is used to collectively define (pseudo-) cubic phases, as they typically exhibit similar optoelectronic properties. However, the black phase is unstable at room temperature. In order to predict the optoelectronic properties of IHPs, density functional theory has been employed to calculate their electronic structures. Considering the relativistic correction and spin-orbit interaction, the following two conclusions were obtained. First, the halide has little effect on the specific energy band structure, except for bandgap values, and second, CsPbX 3 perovskites show a direct band gap, indicating that they have broad application prospects in optoelectronics. In addition, Cs + , as for the organic A cation, has little direct effect on the electronic structure near the band edge, although it can indirectly affect the tilting of the [PbX 6 ] 4− octahedral. Therefore, the excitation and recombination of electrons and excitons are confined within the [PbV 6 ] 4− octahedra (figure 4(a)) [53]. Furthermore, Nicholas et al calculated the effective mass (µ) and exciton binding (R * ) energy of CsPbX 3 perovskites by performing magneto transmission measurements, and the values of these parameters increased with the band-gap energy increasing. In other words, these parameters are related to heavy metal halides. At the same time, the dielectric constant from experiments shows that, when the movement of organic cations is frozen at low temperatures, the dielectric screening mechanism of hybrid and inorganic perovskites is basically the same, which is dominated by the relative motion of atoms within the [PbI 6 ] 4− octahedra cage, as shown in figure 4(b) [54].
Whether a semiconductor material is suitable for a solar energy harvester depends on the following photophysical parameters: band gap, light absorption coefficient, carrier diffusion length. For an ideal single-junction solar cell system, the relationship between E g and its theoretical efficiency limit follows the Shockley-Queisser (SQ) limit. For CsPbX 3 , the light absorption range of the material is determined by its band gap from 1.73-2.31 eV, which is controllable by adjusting the ratio of the halogen, as shown in figure 5(a) [34]. The light  absorption capability of the material as the absorption layer is crucial for highly efficient devices. Generally, the visible light absorption coefficient (α) of the material can be calculated by using the following equation: Here, the extinction coefficient (k) is determined by ellipsometry [55]. Steele et al reported that CsPbBr 3 SC wafer has a visible light absorption coefficient of 9.8 × 10 4 cm −1 (figure 5(b)) [52], ten times higher than that of organic dyes, and slightly higher than that of organic-inorganic hybrid perovskites (MAPbBr 3 SC). Furthermore, Roy et al used scanning photocurrent microscopy and time-resolved microwave conductivity to obtain diffusion lengths of holes and electrons in a series of brominated perovskite-type lead perovskite SCs (APbBr 3 , A = MA + , FA + , Cs + ). It was shown that the diffusion length of holes (L h + D ∼ 10 − 50µm) was significantly longer than that of electrons (L e − D ∼ 1 − 5µm), regardless of the A-type cation. However, the electron diffusion length is more significantly affected by the halogen species than the holes (figure 5(c)) [56].
Some lead-free all-inorganic perovskites have also been developed. Among them, the most common is to replace Pb with another group 14 metal (e.g. Sn and Ge) [58][59][60][61]. However, the stability of Sn 2+ and Ge 2+ -based perovskites in air is not as good as that of Pb 2+ . They are rapidly oxidized to more stable Sn 4+ and Ge 4+ once they are exposed to the air, thus leading to rapid collapse of the 3D perovskite structure. Taking Sn-based perovskite as an example, in order to stabilize the structure, several groups reported Sn-based halide perovskite derivatives with the general composition of Cs 2 SnX 6 , in which Sn exists in a stable +4 oxidation state with good stability in air and moisture [62]. This Cs 2 Sn 2 X 6 is a cubic double perovskite structure through removing half of the octahedral Sn atoms, which are formed by Cs + cation and [SnI 6 ] 2− octahedron as anions, as shown in figure 6(a) [63]. Since Sn vacancy defines the regular sublattice of the double perovskite structure, these compounds are called 'vacancy ordered' double perovskite. This structure shows an abnormally dispersed electronic band structure and direct band gap, and the band gap can be as low as 1.3 eV for Cs 2 SnI 6 . Nonlead perovskite-type compounds with formula unit A 3 B 2 X 9 (A is monovalent cation, B is trivalent metal cation) are also derived with the concept of 'vacancy ordered perovskite'. In the above structure, one third of the octahedral sites is supposed to be empty in order to maintain charge neutrality. This structure is also known as '2D layered perovskite derivative'. Among them, the most common is to replace Pb 2+ with Sb 3+ or Bi 3+ , for example, Cs 3 Sb 2 I 9 and Rb 3 Sb 2 I 8 . Another lead-free perovskite structure is derived from the heterovalent substitution of Pb 2+ by B site cation pairs with oxidation states of +1 and +3, the so-called 'double perovskite' structure based on bismuth and monovalent metals, for example, Cs 2 AgBiBr 6 (figure 6(b)) [63]. At present, both theoretically and experimentally, the feasibility of the above alternatives has been confirmed [64,65].
In the hybrid perovskite photovoltaic community, the facet engineering of perovskite thin films is a new strategy to adjust the film characteristics, such as exquisite control of crystal growth, optoelectronic properties, stability of perovskite materials, types of surface defects and the structure of heterofacets [66,67]. At present, there are also some aspects of facet engineering being carried out in inorganic perovskite systems. Dong et al demonstrated that the preferred (100) of CsPbBr 3 SCs and the polar (111) facet have anisotropic surface-dependent electronic structures, defect formation energies and carrier mobilities [68]. Wang et al confirmed through experiments and theoretical calculations that β-phase CsPbI 3 with a strongly preferred (110) orientation has a lower band gap and is thermodynamically stable compared to the orthogonal α-CsPbI 3 and γ-CsPbI 3 [41]. Compared with traditional materials, researchers currently pay less attention to facet engineering in inorganic perovskite films. Therefore, there is still considerable room for inorganic perovskite facet engineering to improve device performance.

Fabrication of IHP films and IHPSCs
As widely accepted today, the PSC is an n-i-p or p-i-n heterojunction solar cell, where the perovskite is used as an intrinsic absorber sandwiched by two selective contacts, i.e. an electron transporting layer (ETL, n) or a hole transporting layer (HTL, p) [61]. Figure 7(a) and table 1 show some representative architectures of PSCs, including regular structure (n-i-p), inverted structure (p-i-n) and HTL-free carbon electrode structure. Much effort have been devoted to improve the performance of IHPSCs, and the efficiency of all these three types of IHPSCs has been rapidly improved in the past few years, as shown in figure 7(b). More recently, Meng et al promoted the certificated efficiency of n-i-p-type devices to 20.1% by interface engineering and perovskite passivation, and Liu et al fabricated CsPb(I, Br) 3 -based solar cells with 21.8% efficiency by employing Boc-S-4-methoxy-benzyl-L-cysteine as a passivator to suppress halide vacancies and coordinate with undercoordinated Pb 2+ [69,70]. It is worth noting that although the high performance of IHPSCSs so far is basically based on regular structures, some dopants in the HTL used in this structure are detrimental to the stability of the device, which will be discussed in section 4. Compared to the doped organic HTLs, inorganic HTLs, such as NiO, have much less influence on the stability of the inorganic perovskite material and   [81] have been successfully used in the inverted structured PSCs [71]. Moreover, this structure is more compatible with the fabrication technology of perovskite-silicon tandem solar cells [72,73]. Recently, Fu et al improved the efficiency of inverted IHPSCs to 19.84% by introducing 1,4-butanediamine to polish the IHP film [74]. IHPSCs based on carbon counter electrodes (C-IHPSCs) has advantages of low-cost fabrication and superior stability because it avoids the use of HTL and precious metal electrodes [75]. To date, the highest efficiency (16.7%) of C-IHPSCs has been reported by Wang et al [76]. Although [87] Vacuum deposition processing CsPbI 2 Br 11.8% Stabilized PCE of 11.5% (stored in the dark with encapsulations).
[88] the efficiency of the IHPSCs has achieved these exciting results, the reported efficiencies are still far below the SQ theoretical limit, as shown in figure 7(c) [77,78].
Regarding the fabrication methods of the IHPSCs, the preparation of IHPs with high crystalline quality is a prerequisite for high device performance (table 2) [70,77,[82][83][84][85][86][87][88]. Solution processing methods, such as spin coating, is the most widely used approach for IHPs. For the one-step spin coating method, the CsPbX 3 precursor is first dissolved in a suitable solvent with a certain stoichiometric ratio, then spincoated, and finally annealed, as shown in figure 8(a) [39]. Common solvents are dimethylformamide (DMF), dimethyl sulfoxide (DMSO) or a mixture of the two. During the spin coating process, anti-solvent (ATS), such as chlorobenzene, benzene, xylene, toluene, isopropanol (IPA), ethyl acetate and chloroform, can be used to increase the nucleation rate of the precursor solution and thus to improve the morphology and coverage of the film [86]. Compared to hybrid perovskites, the crystallization of IHPs usually requires a higher annealing temperature. Although this one-step method is relatively simple, it is difficult to control the film quality and repeatability of IHPs. Subsequent research found that high-quality absorber films can be obtained by solvent-controlled growth.
Nonetheless, the one-step method is not suitable for CsPb(I, Br) 3 with a high proportion of Br due to the solubility limit of Cs salts, especially CsBr (ca. 0.25 M in DMSO, ca. 0.5 M together with PbBr 2 ) [87,89]. In particular, one thing that has to be addressed is that the performance enhancement of the IHPSCs is attributed to the replacement of PbI 2 with DMAPbI 3 for the one-step spin coating method [44]. As shown in figure 8(b), the two-step method involves a deposition of the PbX 2 layer and then a reaction between PbX 2 and CsX via a multi-step dip coating or spin coating process. Further annealing affords the desired crystal phase. This method is usually used to fabricate CsPbBr 3 and CsPbIBr 2 films because it can overcome the solubility limit in the one-step method.
The IHPs can also be deposited by a vacuum vapor method, such as a dual-source thermal evaporation, as depicted in figure 8(c) [90]. Using the vacuum deposition method, the thickness and composition of the resulting film can be precisely controlled as well as the high repeatability. Hutter et al used time-resolved microwave conductivity technology to study the photoelectric properties of vapor-deposited and spincoated black-phase CsPbI 3 thin films [91]. The results showed that the carrier lifetime of vapor-deposited CsPbI 3 unexpectedly exceeded 10 µs, whereas the carrier lifetime of the spincoated samples was only 0.2 µs. Therefore, the device corresponding to the CsPbI 3 films based on vapor deposition has higher efficiency than that based on the solution method. Chen et al prepared high-quality CsPbBr 3 films by this method and used it for the inorganic-perovskite/organic four-terminal tandem solar cell, obtaining an efficiency of 14.03%, which is the highest efficiency of this structure [90].

Thermal and light stabilities
The unsatisfactory thermal stability of the organic-inorganic hybrid perovskite comes from its volatile organic cations, such as FA + and MA + , which are easily lost at moderate temperature (i.e. 100 • C), eventually leading to the thermal decomposition of the perovskite [17,92]. These problems can be easily solved by replacing the organic cations FA + and MA + with the inorganic cation Cs + . Cache et al explained the difference in thermal stability between MAPbBr 3 and CsPbBr 3 by thermogravimetric analyses (TGA) and confirmed that CsPbBr 3 (weight loss onset ∼570 • C) has much better thermal stability than MAPbBr 3 (weight loss onset ∼220 • C), as shown in figure 9(a) [93]. Moreover, the IHPs exhibit much better operational stabilities under continuous light illumination and electrical biases. Zhao et al demonstrated that the IHPSC based on CsPb(I, Br) 3 can sustain its initial efficiency even after 1500 h maximum power point (MPP) operation (figure 9(b)) [38]. Much better operational stability was found to arise from the suppressed ion migration in the IHPs. CsPb(I, Br) 3 was measured to have an ion migration activation energy of ∼0.45 eV, almost independent of the light illumination. Comparatively, the ion migration activation energy of the hybrid perovskite was significantly reduced under a weak light, as shown in figure 9(c) [38]. This difference in the light illuminationdependent ion migration behavior may come from the different rotation freedoms of A-site cations. The much lower rotation freedom of Cs should be able to help stabilize the lattice structure when under light-induced complicated electron-lattice and light-matter interactions.

Phase stability
Currently, to the best of our knowledge, the IHPSCs with the highest efficiency are based on CsPbI 3 . CsPbI 3 has a relatively narrower band gap compared to that of CsPb(I, Br) 2 , much better material stability compared to CsSn(I. Br) 3 and more mature preparation technology. However, a relatively low tolerance factor of 0.807 makes black-phase CsPbI 3 thermodynamically unstable at room temperature. A feasible way is to substitute an appropriate number of I − ions with smaller size Br − ions or Cl − ions. Sutton et al have for the first time reported enhanced ambient stability of bandgap-tunable CsPb(I x Br 1−x ) 3 (figure 10(a)) compared to CsPbI 3 [36]. Tian et al prepared the IHP films through composition engineering, and the resulting solar cells gained a remarkable long-term operational stability [94]. However, this anion alloying method extends the band gap of the material and narrows the absorption range [34]. At the same time, under light or electron beams, phase segregation may occur when Br − ions are introduced into IHP films [95]. Fortunately, theoretical calculations reported by Yin et al showed that CsPbI 2 Br has a stable alloy phase that results from strong Coulomb interactions in the ionic perovskite lattice [96]. Some relatively stable CsPbI 2 Br perovskites have been reported in experiments [97][98][99].
As mentioned above, partial substitution of I − with Br − induces an increased band gap, which has a negative impact on the efficiency of solar cells. From different perspectives, experimental and theoretical calculation results have shown that B-site doping/alloying strategies have no obvious influence on the band gap of the perovskite material, and can improve the phase stability of IHPs as well. A variety of less toxic inorganic cations, such as Sn 2+ , Ge 2+ , Co 2+ , Mn 2+ , Bi 3+ , Eu 3+ and Nb 5+ have been employed to partially replace B-site ions to tune the tolerance factor or the formation energy, as shown in figure 10(b) [100]. An increase in the formation energy of B-site doped CsPbBr 3 suggests an enhancement of the thermodynamic stability of orthorhombic CsPbBr 3 (figure 10(c)) [100]. Eco-friendly Sn 2+ , which has a slightly smaller size than Pb 2+ , exhibits a potential in adjusting the tolerance factor, thus achieving higher phase stability. However, CsSnI 3 has proved to be unstable under ambient conditions due to the oxidation of Sn 2+ to Sn 4+ within IHP films. Stabilizing the valence state of Sn 2+ in Sn-based IHPs is also the focus of the study of the stability of IHPs. Li et al obtained CsSnIBr 2 thin films with low Sn vacancy by using hypophosphorous acid (HPA) additive. With the assistance of HPA, the formation of Sn 4+ was inhibited in the preparation process of CsSnIBr 2 thin films, thus resulting in superior thermal stability [103]. Germanium (Ge) is recognized as a suitable B-site dopant against phase instability, and a full-coverage native-oxide layer can be formed while using Ge doping [104,105]. In addition, aliovalent B-site doped with Bi 3+ , Eu 3+ , Nb 5+ , etc, is also used to enhance the phase stability of IHPs [106][107][108].
Earlier studies have shown that the phase stability of IHPs is grain size-dependent; the phase stability can be improved with decreasing grain sizes [101]. As shown in figure 10(d), Huang et al found that mixing a small number of sulfobetaine zwitterions in the CsPbI 3 precursor solution can stabilize the α phase of the CsPbI 3 films at room temperature. The interaction between zwitterions and CsPbI 3 hinders the rapid crystallization of CsPbI 3 , which reduces the grain size of CsPbI 3 , thus stabilizing the α phase. On the other hand, it has been reported that α-CsPbI 3 QDs and nanocrystals ranging in size from 4-15 nm are phase-stable for months in ambient air, whereas α-CsPbI 3 nanocrystals ranging in size from 100-200 nm quickly transform from the black phase to the non-perovskite phase. This stability seems to be related to crystal size rather than having an apparent connection with size effects. Thus, in more cases, the most perovskite QDs 10-30 nm in size were reported to be more appropriate as nanocrystals. The dependence of the halide composition on the stability of IHPSCs has also been studied in detail (figure 10(e)) [102]. It is revealed that, as mentioned above, Br doping for CsPbI 3 strongly enhances the stability of the resulting IHP QDs under ambient atmosphere. Furthermore, it is demonstrated that quasi-2D perovskites (figure 10(f)), induced by a judicious amount of phenylethyl ammonium iodide, can remarkably suppress undesirable phase transition [34]. Recently, a 2D-3D all-inorganic Cs 2 PbI 2 Cl 2 -CsPbI 2.5 Br 0.5 perovskite was successfully developed to address the phase instability issue of IHPs [109].
In addition to the common strategies described above, some other effective methods have also been developed to improve the phase stability of IHPs. Li and co-workers reported poly-vinylpyrrolidone (PVP)-induced surface passivation engineering to stabilize cubic-phase CsPbI 3 [110]. Acylamino groups of PVP induced electron cloud density enhancement on the CsPbI 3 surface and reduced the surface energy, which is conducive to stabilize α-CsPbI 3 . Xu et al introduced an ultrathin 2D perovskite on the top of 3D IHPs through the selfassembly method [99]. Ultrathin 2D perovskite can prevent moisture penetration, thus improving the stability of IHPs.

Stability of inorganic PSCs
In this section, we summarize several methods to improve the stability of IHPSCs, such as optimization of bulk materials, charge transport layer, interface engineering, carbon counter electrode and device encapsulations.

Optimization of bulk materials
There is no doubt that improving the crystal quality of IHP is the most effective way to enhance the device stability. Chen et al controlled the α-CsPbI 2 Br crystal growth by a gradient thermal annealing (GTA) method and further optimized the film morphology by using green ATS IPA (figure 11(a)) [86]. By using this GTA-ATS synergistic effect, high-quality perovskite thin films were obtained and as a result, the corresponding devices exhibited excellent moisture and oxygen resistance ability. As shown in figure 11(b), Zhao et al fabricated orthorhombic γ-CsPbI 3 thin film with inherent thermodynamic stability through a simple solution process, and manipulated the formation of a size-dependent phase through a proton transfer reaction by using a small amount of H 2 O [111]. Further theoretical calculation results show that γ-CsPbI 3 with lower surface free energy has better thermodynamic stability than δ-CsPbI 3 when its surface area is greater than 8600 m 2 mol −1 and has equivalent photoelectric performance to α-CsPbI 3 . It has also proved effective by adding an appropriate amount of molten salt additives to the all-inorganic perovskite precursor solution to assist the crystallization of the absorption layer [42,112]. Yu et al developed a ureaammonium thiocyanate molten salt modification strategy to fully release and exploit the coordination activities of SCN − to deposit high-quality CsPbI 3 film for efficient and stable allinorganic solar cells (figure 11(c)) [42]. Zhang et al proposed the use of lead formate (C 2 H 2 O 4 Pb) as molten salt to control the crystallization process. The introduction of this molten salt can not only accelerate the mass transfer process, but also reduce the formation barrier energy that changes the crystallization path, thus forming a high-quality perovskite film that can inhibit the formation of pinholes and cracks (figure 11(d)) [112]. Recently, Cui et al developed a new precursor solution system based on a new room-temperature molten salt as a solvent, dimethylamine acetate (DMAAc), to improve the crystallinity of CsPbI 3 perovskite (figure 11(d)) [46]. In addition, these methods to improve the phase stability of IHPs are also beneficial to the device stability.

Interfacial engineering
The properties of the active layer are surely important to the cell performance of IHPSCs, but by the same token device interfaces cannot be ignored. In recent years, remarkable advancements in IHPSCs have been realized through efficient interface engineering using different interface materials including some halogen salts between charge transport and absorption layers. Interfacial treatment can be realized by vacuum or solution methods, which is one of the main topics of interest for academic and industrial researchers.
Similar to organic-inorganic halide PSCs, IHPSCs also use SnO 2 , TiO 2 , C 60 , ZnO, etc, as ETLs, especially the most widely used TiO 2 and SnO 2 [39,69,113]. Previous studies have shown that SnO 2 has a lower conduction band and higher electron mobility compared to TiO 2 , therefore attracting more attention recently [114,115]. In earlier studies, it was found that the conduction band position of SnO 2 may not perfectly match that of IHPs. Ye et al found that lithium fluoride (LiF) treatment on the ETL SnO 2 can present better energy level alignment in the device and passivate interface defects as well ( figure 12(a)) [116]. Further optimization can also bring better device performance with high light stability and high efficiency (>18%).
As mentioned in section 2.2, choosing suitable HTLs is also the key to fabricate high-performance IHPSCs. Suitable HTLs need to meet several requirements, including high mobility, optimal highest occupied molecular orbital energy level, good solubility and film forming properties and low cost. In earlier studies, HTLs were usually doped to improve their hole mobility, for example, spiro-OMeTAD doped with bis(trifluoromethane)sulfonimide lithium salt (Li-TFSI) [117]. However, severe degradation was reported for IHPSCs due to the existence of dopants (e.g. 4-tert-butyl pyridine and lithium salts) [118,119]. Therefore, some research groups attempted a variety of valuable works, such as using hydrophobic HTL and introducing a hydrophobic layer at the conventional IHPs/HTL interface. As shown in figure 12(b), Tan et al found that the treatment towards CsPbI 3 with phenyltrimethylammonium iodide (PTAI) could significantly improve the phase stability of IHP and the device stability, but would not cause the change in the band gap of light absorber [69]. This positive effect is attributed to the formation of PTAIbased low-dimensional perovskite distributed at the grain boundaries, which not only enhances the phase stability of CsPbI 3 , but also effectively suppresses non-radiative recombination. These results show that the treatment towards the back surface of IHPs with organic halides containing hydrophobic functional groups can improve both the efficiency and stability of the device. Tian et al used an amino-functionalized polymer PN4N as the interface modification layer and an undoped polymer PDCBT as the HTL for CsPb(I, Br) 3 -based solar cells ( figure 12(c)) [96]. The introduction of these two layers can realize high-quality perovskite films with better energy level matching and strong electronic interactions, thus resulting in double interface synergistic passivation of surface defects. In addition, the employment of carbon counter electrodes and encapsulation technology have greatly improved the device stability, as shown in figure 12(d) [120].
In order to obtain highly stable IHPSCs, it is necessary to integrate more passivation methods, such as low-dimensional interface materials, undoped HTLs and encapsulation technologies. Heo et al modified CsPbI 3 perovskite films with oxidized MXene (OMXene) nanoplates based on orthogonal spray coating [121]. This method not only improves the humidity stability of the absorption layer, but also enhances the charge separation at the perovskite/ETL interface. Finally, the PSC module based on this method has good hygrothermal and photoelectric stability, as shown in figure 13(a). Recently, Zhao et al reported more exciting results on the stability of IHPSCs ( figure 13(b)) [109]. This work introduced a 2D Cs 2 PbI 2 Cl 2 coating between the perovskite active layer and the HTL to stabilize the interface and carry out the accelerated degradation of encapsulated IHPSCs under constant light and elevated temperature up to 110 • C. According to the results, it was predicted that the inherent life of the device based on this structure under continuous operation at 35 • C could be 51 000 ± 7000 h (>5 years).

Future perspectives
In summary, IHPSCs have become a research hotspot due to their superior thermal stability and light stability, especially, their suitable top cells for tandem solar cells. Currently, the serious phase instability issue of the IHP, especially for CsPbI 3 , has hampered further development of IHPSCs. In response to this problem, many works have proposed their own solutions, which are divided into the following categories: (1) Self-optimization of IHPs including solvent engineering, halogen site replacement, B doping and alloying, and reducing particle size and dimensions. There is no doubt that the stability of IHPSC mainly depends on the properties of the active layer. By regulating the nucleation and crystallization process of the fully inorganic perovskite absorption layer, a thermodynamically stable phase structure is obtained. (2) Interface optimization including the selection of ETL and HTL materials, introducing a passivation layer at the ETL/perovskite and HTL/perovskite interface. In this section, on the one hand, it is necessary to consider the arrangement of energy bands between each layer; on the other hand, the in situ generation of 2D phases can also suppress the phase transition of the absorption layer at high humidity. In addition, compared to ETL, HTL has a greater impact on the stability of IHPSCs. However, previous studies have shown that devices without HTL layers can significantly improve device efficiency, which still lags behind devices with HTL. At present, some inorganic materials, such as metal oxides, carbon derivatives (carbon, graphene or carbon nanotubes) or other organic materials have been successfully introduced as substitutes for spiro-OMeTAD and have expanded their stability in some cases. (3) The employment of carbon electrodes and encapsulation technology. Carbon electrodes have much better thermal stability compared to metal electrodes, so combining carbon electrodes with fully IHP absorption layers is more conducive to thermally stable devices. In addition, it is necessary to develop an internal and external packaging strategy for IHPSCs, which are sensitive to humidity. Internal packaging can be achieved by introducing interfacial hydrophobic layers, while external packaging can be achieved using insulating encapsulation films that are widely used commercially.
The stabilities of IHPs and IHPSCs are often closely related to their defect properties, especially due to the more complex formation process of IHPs and the significant lattice distortion caused by competition. In addition to the top surface, we also need to pay more attention to the passivation effect of bulk perovskite and bottom surface defects. We also need to explore more interface materials that can withstand high temperatures to buffer the rigid contact, lattice mismatch, and interface stress between the inorganic charge transfer layer substrate and IHPs.
Although IHPSCs have made exciting progress in stability and efficiency, there is still much more room for improvement in the following aspects. First, from a certain perspective, currently most of all inorganic perovskite materials are more sensitive to humidity than hybrid calcium titanium materials, and they are in metastable phases at room temperature. Therefore, there is still a long way to go to improve the phase stability of all inorganic perovskite materials. Second, how to combine IHPSCs with silicon solar cells to fabricate high-performance tandem photovoltaic devices is a research hotspot in the future. Third, we should find new ways to prepare more efficient and stable devices in the future with the aid of theoretical calculations and some more photophysical methods. Finally, the manufacturing technology, as well as the modification of device structures and materials, has to comply with the ease of largescale IHPSC manufacturing.