Inorganic halide perovskite materials and solar cells

Organic-inorganic perovskite solar cells (PSCs) have achieved an inspiring third-party-certificated power conversion efficiency (PCE) of 25.2%, which is comparable with commercialized silicon (Si) and copper indium gallium selenium solar cells. However, their notorious instability, including their deterioration at elevated temperature, is still a serious issue in commercial applications. This thermal instability can be ascribed to the high volatility and reactivity of organic compounds. As a result, solar cells based on inorganic perovskite materials have drawn tremendous attention, owing to their excellent stability against thermal stress. In the last few years, PSCs based on inorganic perovskite materials have seen an astonishing development. In particular, CsPbI 3 and CsPbI 2 Br PSCs demonstrated outstanding PCEs, exceeding 18% and 16%, respectively. In this review, we systematically discuss the properties of inorganic perovskite materials and the device configuration of inorganic PSCs as well as review the progress in PCE and stability. Encouragingly, all-inorganic PSCs, in which all functional layers are inorganic, provide a feasible approach to overcome the thermal instability issue of traditional organic-inorganic PSCs, leading to new perspectives toward commercial production of PSCs.


I. INTRODUCTION
Organic-inorganic halide perovskites have attracted extensive attention due to their outstanding photovoltaic properties, including ease of fabrication, 1-4 tunable bandgap (Eg), 5,6 long carrier diffusion length, 7-9 small exciton binding energy, 10 and large absorption coefficient, 11 which lead to the exceptional power conversion efficiency (PCE) exceeding 25%. 12 However, instability issues of organic-inorganic hybrid perovskite solar cells (PSCs), especially under light soaking and at elevated temperature, have hindered their commercialization. The thermal instability arises from the presence of organic cations in perovskites, such as MA + and FA + cations, which may decompose rapidly at high temperatures. Specifically, the critical temperatures for MAPbI 3 and FAPbI 3 thin films above which degradation takes place are 85 ○ C and 150 ○ C, respectively. 13 Practice has proven that it is advisable to stabilize perovskite materials or discover alternative materials to reach optimal device performance (both efficiency and stability). Inorganic halide perovskite materials solve the thermal instability problem of organic-inorganic PSCs by substituting Cs + for the organic cations in the perovskite precursor solution, which is proved to be a facile but effective method. CsPbX 3 perovskites have been reported as early as 1893, 14 whereas their crystal structure and photoconductive properties were identified in 1958. 15 However, as a novel perovskite absorber employed in photovoltaics and in light emission devices, CsPbX 3 developed rapidly over the past years. [16][17][18][19] Recent results showed that inorganic materials with increasing stability possess potential prospects to produce highly efficient and stable commercial devices. 18,19 PSCs are generally composed of perovskite absorber, electron transfer layer (ETL), hole transfer layer (HTL), and electrode. Solar cells based on inorganic perovskite absorbers are defined as inorganic PSCs (devices in which at least the light absorption layer is inorganic) and all-inorganic PSCs (devices with inorganic materials for all functional layers), which corresponding to types II and IV in Fig. 1 (in which we depict traditional n-i-p structures).

RESEARCH UPDATE
scitation.org/journal/apm FIG. 1. Types of n-i-p structure perovskite solar cells.
We will discuss the pros and cons of two kinds of devices in Sec. III.
In this review, we will try to elucidate the structural, optical, and electrical properties of inorganic halide perovskite materials (CsPbX 3 and other inorganic materials) in Sec. II. Then, the research progress of solar cells based on inorganic perovskite absorbers is summarized in Sec. III. In Sec. IV, the stability issues will be discussed. Finally, the future development for the solar cells based on inorganic perovskite absorbers will be addressed.

Structural properties
There are four phases for CsPbX 3 perovskites, including cubic α-phase (Pm3m), tetragonal β-phase (P4/mbm), and orthorhombic γ-phase (Pbnm) (which are generally known as black phases) as well as nonperovskite δ-phase (Pnma) (which is known as a yellow phase). 20 The different phases exhibit various bond angles (φ) of the Pb-I-Pb configuration in CsPbI 3 , more specifically, the α-phase of CsPbI 3 has a φ of 180 ○ , whereas the φ-values of the β-phase and the γ-phase are approximately 170 ○ and 150 ○ , respectively. [21][22][23] In addition, the α-phase of CsPbI 3 can be converted into the δ-phase through slow cooling in the absence of ambient moisture, whereas a tetragonal β-phase CsPbI 3 will be formed from a γ-phase CsPbI 3 under rapid cooling conditions, as shown in Fig. 2(a). 21 The phase change with respect to temperature is one of the serious instability issues in the inorganic perovskites. Moreover, the crystal structures for CsPbCl 3 and CsPbBr 3 will convert into a pure cubic phase from their tetragonally distorted phase at 47 ○ C and 130 ○ C, respectively. 15 For CsPbI 3 , the phase transition point from the orthorhombic phase to the cubic perovskite phase is at 305 ○ C. Unfortunately, the cubic perovskite phase of CsPbI 3 is not stable under ambient conditions, and it thus quickly converts back to the orthorhombic phase. This is a reversible process.
Recent reports have proposed that the Goldschmidt tolerance factor (t) and octahedral factor (μ) can be used to predict the stability of a perovskite structure. Sun and Yin illustrated this by presenting a map of (t, μ) for 138 perovskite compounds, as reproduced in Fig. 2(b). 24 Li et al. concluded that perovskites were stable when 0.813 < t < 1.107 and 0.377 < μ < 0.895, and the t and μ are given by the following expressions: RESEARCH UPDATE scitation.org/journal/apm μ = r B r X , (2) where r A , r B , and r X are the ionic radii of ions A, B, and X, respectively. The t values of CsPbI 3 and CsPbBr 3 are 0.81 and 0.814, and the corresponding μ values are 0.54 and 0.62, respectively. Importantly, the Cs + cation is too small to form a stable perovskite phase of CsPbI 3 . Hence, tailoring of the halogen ratio is a feasible method to increase the stability of I-rich inorganic perovskite films [I − , (2.2 Å), Br − (1.96 Å), and Cl − (1.81 Å)]. 25

Electrical properties
The bandgaps of CsPbX 3 perovskites have been found to change when the halogen type is adjusted. CsPbI 3 , CsPbI 2 Br, CsPbIBr 2 , and CsPbBr 3 have bandgaps of 1.73 eV, 26 1.92 eV, 27 2.05 eV, 28 and 2.3 eV, 29 respectively. Obviously, CsPbI 3 , with a more suitable bandgap, has the potential to yield higher efficiency when compared to CsPbBr 3 in a single junction device. However, the phase instability of CsPbI 3 limits its further progress. Meanwhile, CsPbBr 3 with a bandgap as desired for tandem or spectral splitting systems exhibits a relatively good stability under electron beam irradiation, constant illumination, air exposure, and thermal stress. 30 Interestingly, CsPbIxBr 3−x perovskite absorbers have great potential to obtain efficient as well as stable PSCs.
For a long time, the unique dipolar nature of molecular cation has been considered as a source of slow light-induced charge carrier recombination rates. [31][32][33][34][35][36] This misconception was not resolved until Fafarman and co-workers performed transient terahertz spectroscopy on inorganic CsPbI 3 thin films. 37 The charge carrier mobility greater than 30 cm 2 V −1 s −1 and a recombination rate of 10 −10 cm 3 s −1 for CsPbI 3 suggest that avoidance of the organic cation is not the only factor leading to an extremely low recombination rate. A low recombination rate is not uniquely due to the presence of a molecular dipole. Savenije et al. proposed that the charge carrier mobility and recombination rate are primarily based on the inorganic octahedral network rather than on the organic nature of the A-site cation, which confirms the above statement. 38 Thus, Csbased inorganic perovskite materials possess the desirable electrical properties needed for high efficiency devices.
Nicholas et al. obtained the transmission spectra of CsPbX 3 (CsPbBr 3 , CsPbI 2 Br, and CsPbI 3 ) over a wide range of temperatures (from 4.2 K to 270 K). A blue shift consistent with a broadening band-edge absorption energy is observed with increasing temperature. The bandgaps of inorganic materials clearly exhibit a wellbehaved monotonic dependence on temperature. On the other hand, the bandgaps of organic-inorganic materials increase with increasing temperature because of occurring phase transitions, 39 which suggests that the overall electronic properties of halide perovskites are mainly dependent on the presence of a molecular dipole. 40 Nicholas et al. also accurately measured the exciton binding energy (R * ) of MAPbI 3 directly, using extremely high magnetic fields, obtaining a precise value of 16 ± 2 meV, 39 which is less than earlier values of 37-50 meV. 41,42 Since then, they used the same principle to determine the R * and reduced mass (μ) of inorganic CsPbX 3 perovskites. The R * values of CsPbI 3 , CsPbI 2 Br, and CsPbBr 3 are 15 ± 1 meV, 22 ± 3 meV, and 33 ± 1 meV, and the corresponding μ values are 0.114 ± 0.01 m 0 , 0.124 ± 0.02 m 0 , and 0.126 ± 0.01 m 0, respectively. 40 The effective dielectric constant values (ε eff ) of CsPbI 3 (10), CsPbI 2 Br (8.6), and CsPbBr 3 (7.3) have also been acquired by calculation from the values of R * and μ. The value of ε eff is found via the following equation: where R 0 is the atomic Rydberg constant and m 0 is the free electron mass. Ion migration is yet another drawback in organic-inorganic PSCs. Zhao et al. demonstrated that light-induced ion migration can be eliminated by substitution by inorganic Cs + . 43 Moreover, they adopted a type of poling process for CsPbI 2 Br and MAPbI 3 films under various light intensities to acquire an ion migration barrier (Ea). The Ea for MAPbI 3 decreased from 0.62 eV under dark conditions to 0.07 eV with a change in light intensity from 0.1 to 25 mW/cm 2 , which was in stark contrast with CsPbI 2 Br, for which the Ea merely reduced to 0.43 eV from 0.45 eV. The Ea values derive from the following equations: The σ ion was obtained by subtracting the electronic conductivity σ electron , which was galvanostatically measured, from σ total .

B. Other inorganic perovskite absorbers
To solve the problem of lead toxicity, an increasing number of researchers have focused on the substitution of lead-free elements in organic-inorganic perovskites as well as in inorganic PSCs. Similar to CsPbX 3 PSCs, the bandgaps of CsSnX 3 are also dependent on the halogen element ratio, varying from 1.23 eV to 1.75 eV by tuning the mixture of X elements. The bandgaps of CsSnI 3 , CsSnI 2 Br, CsSnIBr 2 , and CsSnBr 3 are 1.27 eV, 1.37 eV, 1.65 eV, and 1.75 eV, respectively. The bandgaps of CsGeI 3 , CsGeBr 3 , and CsGeCl 3 are approximately 1.53 eV, 2.32 eV, and 3.67 eV, respectively. 44 Oxidation of Sn 2+ to Sn 4+ ions in ambient atmosphere is commonly observed for Sn-based perovskites. To solve this problem, A 2 SnX 6 (2-1-6) as perovskite derived material emerged as a new crystal structure. 45 In addition, a number of other structures, such as A 3 B 2 3+ X 9 (3-2-9) and A 2 B 1+ B 3+ X 6 (2-1-1-6), deviate from ABX 3 . The A 2 BX 6 structure equivalently removes half of the B-site cations of the ABX 3 perovskite checkerboard patterns, where the B-site cation in this structure is in a +4 oxidation state. Concomitantly, the A 3 B 2 X 9 structure likewise removes one third of the B-site cations from the ABX 3 perovskite crystal lattice. Moreover, the A 2 B 1+ B 3+ X 6 structure is a so-called novel "double perovskite" in which the B 2+ cation is replaced by metal cations with different valence states including B 1+ and B 3+ . 46

III. SOLAR CELLS
In this section, the research progress in solar cells based on inorganic perovskite absorbers is described both for inorganic PSCs and all-inorganic PSCs. To clearly summarize the research progresses and performance properties of solar cells based on inorganic materials, detailed performance parameters of inorganic PSCs and all-inorganic PSCs with different compositions are shown in Tables I  and II, respectively.

RESEARCH UPDATE
scitation.org/journal/apm  A. Research progresses of inorganic PSCs
The reproducibility of high-quality CsPbBr 3 films is one of the challenges in the sequential deposition processes described above. To study this, Liu et al. deposited CsPbBr 3 perovskite films by dualsource coevaporation of CsBr and PbBr 2 precursors. 52 Ultimately, they achieved a stable PCE of 6.95% for 0.09 cm 2 and 5.37% for 1 cm 2 through optimization of the substrate temperature, the ratio of evaporation rates, in combination with the postannealing temperature. Furthermore, Li et al. also prepared perovskite films by using the dual-source vacuum coevaporation method [ Fig. 3 with stoichiometric precursors of CsBr and PbBr 2 , which yielded devices with an ultrahigh open-circuit voltage (Voc) of 1.44 V and a PCE of 7.78%. 53 Meanwhile, they pointed out that the CsPb 2 Br5 phase appeared when the precursor ratio was PbBr 2 :CsBr = 1:0.8. Jiang et al. reported the advantages of CsPb 2 Br5 in which the perovskite grain size was enlarged along with effective passivation of grain defects. In addition, the energy barrier for ion migration was increased and the carrier recombination rate was reduced when introducing CsPb 2 Br5 at the perovskite/HTL interface. 54 The schematic structure and energy level diagrams of CsPb 2 Br5-based devices are depicted in Fig. 3

RESEARCH UPDATE scitation.org/journal/apm
characteristically is a strongly reducing agent and a strong base, which potentially suppresses the oxidation of Sn 2+ . Moreover, a high temperature removal of redundant hydrazine was not required in this work due to the high volatility of this agent. A PCE of a device prepared by the reducing hydrazine vapor atmosphere process up to 3.04% was obtained, which is much higher than that of organic-inorganic Sn-based PSCs (about 1.02%). To improve the device performance, numerous investigations have been done on electrodes, 58,59 ion modification, 60,61 and the matching of energy levels. 51,[62][63][64] These works based on all-inorganic PSCs will be described below in detail.

Triiodo-based inorganic PSCs
CsPbBr 3 with a rather high Eg (2.3 eV) is not likely to be an efficient absorber in single junction solar cells, 65 although it is rather stable. However, replacing Br of CsPbBr 3 entirely by I can be very effective in reducing the bandgap. As mentioned above, the bandgap of CsPbI 3 is 1.73 eV, which is the most suitable bandgap among the CsPbX 3 perovskites for high-efficiency solar cells.
The first CsPbI 3 PSC was reported by Choi et al., showing an insufficient PCE of 0.09% on account of an unstable cubic crystal structure. 16 The structural properties of CsPbI 3 are depicted in Fig. 4(a). To solve this issue, Snaith et al. added HI to the perovskite precursor solution. This method was widely used in later studies. PbI 2 and HI reduce the bandgap of perovskite from 1.72 eV to 1.68 eV under the influence of tensile lattice strain. 66 The HI plays a critical role in reducing the phase transition temperature and inducing smaller grains other than creating lattice strain. It is worth noting that the small grain size is critical for stabilization of the black phase, as evidenced by the result of Snaith. 67 A 2.9% efficiency was achieved with significant hysteresis; therefore, the stabilized PCE was just 1.7% in this work. Hayase et al. incorporated a MoO 3 layer with a thickness of 3 nm as a hole extraction layer before the deposition of the Au electrode to suppress the I-V hysteresis in the structure of FTO/b-TiO 2 /m-TiO 2 /CsPbI 3 /P3HT/Au. 68 Finally, they obtained a forward sweep efficiency up to 4.68% and a reverse sweep efficiency up to 3.78%. In addition, Luo et al. introduced a novel intermediate phase of Cs 4 PbI 6 to stabilize the α-phase of CsPbI 3 and obtained 4.13% efficiency. 69 Savenije et al. reported vapor-deposited CsPbI 3 films with a charge carrier mobility of 25 cm 2 /(V s) and a charge carrier lifetime of 10 μs, resulting in a PCE approaching 9%. 38 Lin et al. also adopted a vacuum-deposited method to form perovskite films and achieved approximately 10% PCE, through optimizing the precursor ratio and annealing time. 70 In order to increase the PCE beyond 10%, Luther and coworkers fabricated stable CsPbI 3 quantum dot (QD) films. 26 To overcome the low carrier transport efficiency of QD films, Liu et al. introduced a μ-graphene (μGR) with various functional groups (hydroxyl, carbonyl, and carboxyl) into the QD films to form a crosslink structure, as shown in Fig. 4(b), which exhibited a high carrier transport efficiency and conductivity. 71 Hence, the PCE increased to 11.4%, which achieved a 12% increase compared to the reference cell. For the same purpose, Luther et al. dissolved A-site cation-supplying halide salt (FAI, FABr, MAI, MABr, and CsI) in ethyl acetate (EtOAc) solution to treat CsPbI 3 QDs films, aiming to tune the coupling between perovskite QDs. Here, the CsPbI 3 QDs films were prepared by layer-by-layer technology. A saturated Pb(NO 3 ) 2 solution in methyl acetate was used to soak the CsPbI 3 QDs films after the formation of each QD layer and to remove the native ligands that are harmful to charge transport. Ultimately, the FAI-coated device displayed significant efficiency enhancement, increasing to 13.43%, from an initial value of 8.5%. 72 Yuan et al. have fabricated CsPbI 3 QD solar cells with a series of dopant-free polymeric HTMs [2,2 ′ ,7,7 ′ -Tetrakis(N,Ndi-p-methoxyphenylamine)-9,9 ′ -spirobifluorene (spiro-OMeTAD), P3HT, PTB7, and PTB7-Th], where the device with PTB7 achieved a remarkable PCE of 12.55% with a Voc of 1.28 V together with an extremely low energy loss of 0.45 eV. 73 The phenylethylammonium iodide (PEAI) with a large cation has been widely applied to organic-inorganic PSCs, realizing a

RESEARCH UPDATE
scitation.org/journal/apm two-dimension (2D) Ruddlesden-Popper structure that enhances the stability of perovskite devices. [74][75][76] Many groups have also introduced PEAI into CsPbI 3 perovskite films to improve the PCE and phase stability of devices. The optimal Cs 0.9 PEA 0.1 PbI 3 device by Liu et al. yielded an efficiency of 5.7%. 77 The optimal PEA 2 Cs n−1 PbnX 3n+1 (n = 60) device achieved an efficiency of 12.4%, as reported by the group of Yuan [the change in perovskite structure and decomposition energetics with varying n values is shown in Fig. 5(a)]. 78 Zhao et al. synthesized EDAI 2 by using EDA and HI, and subsequently, EDAPbI 4 was synthesized by employing EDAI 2 and PbI 2 . They controlled the amounts of EDAPbI 4 in CsPbI 3 to form effective low-dimensional CsPbI 3 ⋅xEDAPbI 4 films and achieved a champion efficiency of 11.86% when x = 0.025. 79 Considering the quantum confinement effect of quasi-two-dimensional perovskites, they adopted PEAI to form a 2D capping layer on top of the 3D perovskite surface to improve the short-circuit current density (Jsc) of the devices [ Fig. 5(b)]. The as-prepared PEA + -CsPbI 3 -based PSC exhibited 14.3% efficiency with a very high Jsc of 18.5 mA/cm 2 , which increased 21% compared to that of a pure CsPbI 3 -based perovskite device. 80 In addition to a low-dimensional mixed perovskite, a pure low-dimensional perovskite has also been reported by the group of Kuan. 81 They fabricated pure 2D cesium lead iodide perovskite, BA 2 CsPb 2 I7 [the XRD pattern and structural diagram are shown in Fig. 5(c)], showing comparatively a low PCE of 4.84%, as a result of quantum confinement effect of low dimensional perovskites.

RESEARCH UPDATE scitation.org/journal/apm
Increasing the Goldschmidt tolerance factor through ion embedding is another way to improve the PCE and stability of perovskites. Miyasaka et al. incorporated Eu 3+ (EuCl 3 ) into CsPbI 3 and achieved an efficiency of 6.8%, although Eu 3+ occupies only few interstitial sites rather than replacing Pb 2+ , as verified from the shifts of XRD peaks. The XRD peaks shifted to lower 2θ with increasing content of Eu 3+ , but Eu 3+ possesses smaller ionic radii (0.95 Å) than Pb 2+ (1.19 Å). 82 Bi 3+ (1.03 Å) 83 and Ca 2+ (1.00 Å) 84 were introduced into the perovskite to replace Pb 2+ , thus forming CsPb 0.96 Bi 0.04 I 3 and CsPb 0.95 Ca 0.05 I 3 , respectively. The Goldschmidt tolerance factor of perovskite crystals increases with decreasing substitutive B-site ion radius. Complete substitution has also been studied by these researchers, apart from partial substitution of Pb 2+ , which reduces both tolerance factors and lead content. The metal cations Sn 2+ (1.02 Å) and Ge 2+ (0.87 Å), which have ns 2 lone pairs, are considered to be the most promising candidates for replacing Pb 2+ because they both belong to the same group (IV-A group) in the periodic table. Nevertheless, the inherent defects formed by the Sn-cation vacancies in CsSnI 3 films result in a decrease in the conductivity of the films. 85 To reverse this situation, SnF 2 , SnCl 2 , SnBr 2 , and SnI 2 were introduced into the perovskite to reduce the formation of Sn vacancies. The PCE was improved from an almost negligible efficiency to 2.02% by adding a 20 mol. % of SnF 2 to the perovskite precursor solution. 86 SnF 2 never disturbs the lattice structure of CsSnI 3 but fills the vacancies that are generated by Sn. The excess SnI 2 in the CsSnI 3 solution functions as both the compensator and suppressor of Sn 2+ vacancies. 87,88 Hatton et al. compared the functions of SnF 2 , SnCl 2 , and SnBr 2 together with SnI 2 additives in CsSnI 3 cells. Here, SnCl 2 was more effective in terms of improving efficiency and stability [the SEM image of films with and without SnCl 2 additives is shown in Fig. 6(a)]. 89 However, finally, only 3.56% efficiency was obtained, which is considerably lower than that of Pb-based inorganic solar cells. Chen et al. prepared CsSn 0.5 Ge 0.5 I 3 , which exhibited a striking efficiency of 7.11% and excellent stability, using a one-step vaporprocessing method. 90 The CsSn 0.5 Ge 0.5 I 3 raw powder was synthesized by powder precursors of CsI:SnI 2 :GeI 2 . Huang et al. added sulfobetaine zwitterions, which interact with PbI 2 ⋅DMSO colloids, to control the crystallization of perovskite and decrease the colloid size. 91 The cell exhibited an efficiency as high as 11.4%.
Recently, orthorhombic black γ-phase CsPbI 3 with lower surface free energy develops rapidly owing to their more suitable bandgaps (1.68 eV) and sufficient stability. 17,92-95 Hu et al. added a small quantity of H 2 O into the CsPbI 3 precursor solution to achieve a stable perovskite film at room temperature and achieved a PCE of 11.3%. PbI 2 colloids, a molecular complex DMF⋅xHI, as well as Cs + and I − ions were mixed in the precursor solution with H 2 O. The H 2 O-induced proton transfer process caused an increased activity of dissociative H + and I − and gradually splitted PbI 2 colloids, which reduced the size of the colloids, in order to form the corner-sharing [PbI 6 ] 4− octahedra intercalated by Cs + during spincoating. After thermal annealing at 100 ○ C, perovskite-structured polymorphs with decreased crystallite size were formed, stabilizing the γ-CsPbI 3 films with increased surface area. The scheme for the stabilization of γ-phase CsPbI 3 is shown in Fig. 6(b). 93  In addition, they employed upconversion nanoparticles (UCNPs) to active PTAA for an enhanced utilization of sunlight and prepared nitrogen-doped graphene quantum dots (N-GQDs) on top of the SiO 2 substrate to convert damaging UV light into useful visible photons to improve Jsc, thereby increasing the PCE of γ-phase CsPbI 3 over 15%. 94,95 They also realized an encouraging efficiency of 16.07% by doping chlorine in a black γ-CsPbI 3 precursor solution. 17 In addition, a slightly distorted β-phase CsPbI 3 has better stability than α-phase CsPbI 3 . However, there are few studies in this direction. Fu et al. fabricated β-phase CsPbI 3 by adding PEA acetate in perovskite precursor solution and generated a PCE of only 6.5%. 96 Surface modification is another common method for markedly upgrading the performance of inorganic PSCs. Zhu et al. introduced DETA 3+ with oil-wet hydrocarbon chains to enhance the hydrophobicity of perovskites and simultaneously inhibit the formation of large grains. 97 The polymer poly-vinylpyrrolidone (PVP) utilized as a passivator possesses a lower surface energy, which originates from acylamino groups of PVP. 98

I/Br mixed I-PSCs
CsPbI 3 perovskites have achieved an efficiency of 18.4% with greatly improved stability as compared to the previously discussed PSCs. However, a rather low relative humidity (RH) still needs to be maintained in the operating environment. To overcome the issues of a too wide bandgap of CsPbBr 3 and instability of CsPbI 3 , researchers tuned the I/Br ratio and formed I/Br-mixed CsPbIxBr 3−x . The most commonly used perovskites are CsPbIBr 2 and CsPbI 2 Br. CsPbIBr 2 shows relatively slow development as compared to CsPbI 2 Br, though the development of both started in 2016. Ho-Baillie et al. was the first group to use spray-assisted solution-processing to produce an inorganic CsPbIBr 2 PSC and obtained an efficiency of 6.3% with negligible hysteresis. 101 Cheng et al. reported the CsPbIBr 2 PSC and obtained a stabilized efficiency of 6.07%, a reverse sweep efficiency of 8.02%, and a forward sweep efficiency of 4.02% with clear hysteresis. They explained that "I-rich" CsPbI (1+x) Br (2−x) phases segregated and clustered at grain boundaries under light and electron beam illumination, resulting in hysteresis. 102 To avoid high temperatures during the fabrication process and low electron mobility of TiO 2 in traditional n-i-p structures, Zang et al. deposited In 2 S 3 films as electron transfer layers with a low-temperature fabrication process. Unfortunately, the PCE of 5.59% was still low. 103 To overcome the significant recombination present in CsPbIBr 2 perovskite films, SmBr 3 was introduced to generate an energy gradient, which modified the TiO 2 surface and led to a perfect ETL/perovskite interface. The PCE of SmBr 3 -modified CsPbIBr 2 PSC was up to 10.88%, an increase by 30% compared to the reference cell. Under this condition, charge recombination and nonradiative recombination were suppressed as well as charge extraction was improved. 104 Due to the wide bandgap (2.05 eV) of CsPbIBr 2 perovskites, substituting Pb with Sn is an effective method to reduce the bandgap and improve the PCE. The optimal composition, CsPb 0.75 Sn 0.25 IBr 2 (1.78 eV), was achieved via regulating a series of CsPb 1−x SnxIBr 2 perovskites. This inorganic perovskite device with an impressive efficiency of 11.53% and an Voc of 1.21 V will be a potential highefficiency candidate owing to its much better light stability and phase stability. 105 However, the fully Sn-based CsSnIBr 2 perovskite has a relatively low efficiency of 1.56%. 106 Substituting Pb with Sn is not a perfect method to enhance efficiency; it appears to be a more feasible approach to further tune the I/Br ratio to form CsPbI 2 Br perovskites. The first CsPbI 2 Br PSC has shown a surprisingly high PCE of 9.84%. 27 Recently, the PCE of CsPbI 2 Br PSCs has shown rapid progress to 16.2% as a result of dual interfacial design. 107 Most researchers obtained vacuum-deposited CsPbI 2 Br films with various stoichiometric ratios of CsBr:PbI 2 . In this way, the best efficiency of devices has been achieved at a stoichiometric ratio of 1:1 (CsBr:PbI 2 ). 70,108 Interestingly, CsBr-rich CsPbI 2 Br presented relatively good stability upon air exposure, as Ho-Baillie et al. reported.

RESEARCH UPDATE scitation.org/journal/apm
The efficiency of a device prepared via the antisolvent assisted crystallization method of 16.07% was reported by Li et al. 18 A highquality perovskite film with reduced defect density was formed by using a gradient thermal annealing method. The morphology of the CsPbI 2 Br film was controlled by adopting a green antisolvent isopropanol to control the growth of α-phase. A schematic illustration of the CsPbI 2 Br perovskite crystallization process is shown in Fig. 7(a). Although the antisolvent assisted crystallization method can lower the annealing temperature, it is difficult to induce crystallization at temperatures below 150 ○ C. Zhao et al. replaced PbI 2 with HPbI 3+x and found an optimized annealing temperature of 130 ○ C; this method has also been applied in preparing CsPbI 3 . 109 Ion incorporation technology, including A-, B-, and X-site substitutes and interstice filling, has been widely used to optimize device performance in recent years. Potassium cations (K + ) were introduced into CsPbI 2 Br to form the Cs 0.925 K 0.075 PbI 2 Br perovskite and a PCE of 10.0% was obtained. 110 Germanium cations (Ge 2+ ), 111 strontium cations (Sr 2+ ), 112 and europium cations (Eu 2+ ) 113 were incorporated into CsPbI 2 Br to upgrade PCE to 10.8%, 11.2%, and 13.71%, respectively. They all substituted Pb 2+ ions to reconstruct a new host lattice, thus yielding CsPb 0.8 Ge 0.2 Br, CsPb 0.98 Sr 0.02 I 2 Br, and CsPb 0.95 Eu 0.05 I 2 Br, respectively. The group of Yin for the first time introduced the smaller element fluorine (F) into CsPbI 2 Br to form a novel perovskite structure, CsPbBrI 2−x Fx. 114 The Goldschmidt tolerance factor was enlarged by partially substituting I by F; thus, the α-phase CsPbI 2 Br structure became more stable. The CsPbBrI 1.78 F 0.22 PSC displayed a superior structural stability than CsPbI 2 Br PSC when all devices were stored under 20% RH at room   115 The molecule copper(II) bromide (CuBr 2 ) was also incorporated into the CsPbI 2 Br precursor to control the crystallization of perovskite by suppressing the nucleation rate during the annealing process, resulting in a high quality all-inorganic perovskite film. Thus, a PCE of 16.15% was obtained, which is close to the maximum efficiency of CsPbI 2 Br PSCs. 116 Suitable band energy alignment has always been a prerequisite for achieving high Voc in perovskite devices, owing to the presence of an undesirable energy barrier in the electronic band structure, which may result in considerable charge recombination, thereby limiting the quasi-Fermi level splitting. 117 The group of Liu successively adopted a 3D-2D-0D dimension-profiled interface structure, CsPbI 2 Br-CsPbI 3 QD component structure, and FA + modified QD component to arrange a graded energy alignment. Figure 7(b) exhibits the variation of energy levels with different dimensions (3D, 2D, and 0D) of structures at the graded heterojunction interface. 118 In this case, a high Voc of 1.19 V was realized, which originated from an enhancement of the built-in electric field. The CsPbI 2 Br/CsPbI 3 QD interface showed a comparable graded bandgap. The device achieved a PCE as high as 14.45%, exhibiting significant improvements compared to individual layer devices, especially of Voc. 119 Regarding the structure of CsPbI 2 Br/FA + -CsPbX 3 , an FA + -CsPbX 3 layer generated by FA + modifying cations could passivate the grain boundary and adjust the energy level. Simultaneously, the FA + modified QD layer plays a significant role in adjusting the band edge bending as well as in reducing the surface defects. 120 Except for energy level arrangement optimization of the perovskite layer, the energy level alignment of the charge transport layer is equally crucial. Yip et al. replaced the SnO 2 layer with the SnO 2 /ZnO bilayer ETL; the corresponding energy diagrams are shown in Fig. 7(c). The Voc was improved to a value as high as 1.23 V, from 1.06 V, and the device efficiency increased to a desirable value of 14.6%, from 11.9%. 121 Interface passivation has always been a viable method to improve the performance of solar cells, and CsPbI 2 Br is no exception. As one of the commonly used hole transport materials, P3HT also passivates the perovskite film through a postannealing process, which is ascribed to a reduction in electron-hole recombination by the polythiophene in P3HT. As shown in Fig. 7(d), steady-state PL indicated that P3HT passivates the surface trap states of CsPbI 2 Br, thus increasing the average charge carrier lifetime of CsPbI 2 Br films by a factor two (from 6.91 to 14.8 ns). The Voc value of CsPbI 2 Br with passivating P3HT was up to 1.32 V with an E loss of 0.5 V, which implies an increase of almost 0.14 V over that of spiro-OMeTADbased devices. 122 An Voc of 1.31 V was also obtained using the hole transport material PIF8-TAA by the group of Im. 123 Tian et al. introduced Pb 2+ derived from Pb(NO 3 ) on the perovskite surface by means of a postprocessing strategy to reduce defect states resulting from Pb vacancies and I interstitials. 124 The passivation effects of PEAI, PEABr, and PEACl on perovskite films were compared by Liu et al. Results showed that they all have improved film stability without compromising efficiency, where the PEAClbased film presented rather good stability compared to PEAI and PEABr-based films. The PEACl-based film did not show any change when exposing the films to air with a RH of 40%-50%, and they still exhibited excellent moisture stability at a humidity even beyond 60% RH, which was attributed to PEA + -passivation of surface defects and Cl − induced lattice shrinkage. 125 Recently, the concept of dual interfacial design was presented, in which the polymer (PN4N) with amino-functionality and PDCBT without dopant acted as a cathode interlayer and an anode interlayer, respectively. The PN4N plays a large role in reducing the work function of SnO 2 through the action of an interfacial dipole and in tuning the surface wetting property of SnO 2 . Meanwhile, the PDCBT with a deeper HOMO level provided a better energy alignment. 107 The cell with dual interfacial design achieved an encouraging efficiency of 16.2%, which is a record efficiency of CsPbI 2 Br as far as known.

Solar cells based on perovskite derived materials
The Cs 2 SnI 6 composite, a material derived from perovskite, has been explored as a structure to stabilize Sn-based perovskites because Sn in this structure exists as a +4 oxidation state. However, no devices have been made with Cs 2 SnI 6 as the absorber with an efficiency of more than 1%. 126,127 Therefore, Cs 2 TiBr 6 is preferred, with a favorable bandgap of 1.8 eV, as fabricated by vapor-based deposition of CsBr and TiBr 4 successively. Devices with this absorber material showed a PCE of 3.28%. 128 The corresponding SEM images of the surface morphologies of the thin films and the formation mechanism of the Cs 2 TiBr 6 thin films are shown in Fig. 8(a). Subsequently, a solar cell based on Cs 3 Bi 2 I 9 has been reported by Johansson et al.; a schematic diagram of the Cs 3 Bi 2 I 9 structure is depicted in Fig. 8(b). They first employed P3HT as HTM in the structure FTO/c-TiO 2 /m-TiO 2 /Cs 3 Bi 2 I 9 /P3HT/Ag, which led to 0.4% PCE. 129 Then, spiro-OMETAD was used to replace P3HT, and this device yielded a slightly higher PCE of 1.09%. 130 A Rb 3 Sb 2 I 9 solar cell device with the same 3-2-9 structure reached an efficiency of 0.66%. 131 Double metal Cs 2 AgBiBr 6 PSC [crystal structure is shown in Fig. 8(c)] by replacing two Pb 2+ with Ag + and Bi 3+ in the crystal lattice has gained public attention, owing to the convenient substitution of its chemical composition. Many research groups have reported on it, both p-i-n structures 132 and n-i-p structures. [133][134][135] Nevertheless, the maximum efficiency remains below 3%. The underlying cause can be ascribed to the underdeveloped electronic structure, materials properties, film quality of the perovskite-derived material, and device architecture. It is difficult to develop a synthetic route to obtain uniform thin films of the correct phase and composition. Another important reason is the wider bandgap of perovskite derived materials, such as Cs 2 AgBiBr 6 and Cs 2 AgBiCl 6 , which have bandgaps of 2.19 eV and 2.77 eV, respectively. 136 Various strategies have been attempted to engineer the bandgap and to modify the device structures in order to improve the device performance, such as chemical doping and alloying. Sometimes they exhibited promising potential for other optoelectronic applications. 137 Overall, it is still a long way to enhance the PCE of perovskite-derived materials and solar cells based on them.

B. Research progress of all-inorganic PSCs
Owing to the excellent stability of CsPbBr 3 thin films, the carbon-based device structure, which do not have a hole transport layer, becomes the primary choice for CsPbBr 3 -based devices. This stems from the instability of the organic hole transport layer and the metal electrodes, which are prone to react with the halide ions in the perovskite layer. In 2016, FTO/c-TiO 2 /m-TiO 2 /CsPbBr 3 /carbon structure-based solar cells were reported almost simultaneously by Chen et al. and Liu et al., and they obtained a PCE of 5% 138 and a slightly higher efficiency of 6.7%, 29 respectively. In order to reduce the interface charge recombination between the perovskite and carbon electrode, the group of Tang prepared electrodes by varying the ratio of multi-walled carbon nanotubes (MWCNT) to carbon black (CB) 59 or alternatively, by inserting PtNi nanowires in carbon ink, 139 respectively. Except from the improvements in the electrodes, intermediate layers were inserted in ETL/CsPbBr 3 and CsPbBr 3 /carbon interfaces to speed up charge extraction. Carbon QDs, 63 CuInS 2 /ZnS QDs, 140 and graphene QDs 51 can be used as intermediate energy levels at the ETL/CsPbBr 3 interface, while red phosphorus QDs can be used as intermediate energy levels at CsPbBr 3 /carbon interfaces. The graphene QD modified CsPbBr 3 device achieved a PCE of 9.72%, supreme value among the PSCs with intermediate energy level regulation. 51 Likewise, ion incorporation engineering in all-inorganic PSCs has drawn tremendous attention. Alkali metal cations were used as dopants in CsPbBr 3 , partially replacing Cs + , which resulted in suppressed nonradiative losses and radiative recombination. The asprepared Rb + -doped Cs 0.91 Rb 0.09 PbBr 3 perovskites achieved a PCE as high as 9.86%, benefitting from the changed lattice structure and optimized energy levels. 61 A series of lanthanide ions were separately incorporated into the perovskite lattice to substitute Pb + , aiming to enlarge grain size and prolong carrier lifetimes, where the performance of all-inorganic CsPb 0.97 Sm 0.03 Br 3 solar cells was significantly enhanced. A gratifyingly high efficiency of up to 10.14% was reached, and an ultrahigh Voc of 1.594 V was realized. 60 A SEM image of the cross section and the J-V curve are shown in Fig. 9(a).
There are a few structures based on all-inorganic CsPbI 3 PSCs. Chen et al. achieved a PCE of 5.18% by doping CsPbI 3 with Sb to form CsPb 0.96 Sb 0.04 I 3 . 141 They employed carbon as an electrode and the resulting device is without HTM. Again, they replaced PbI 2 in the perovskite solution with HPbI 3 to improve device efficiency and stability based on the above HTL-free carbon-based device structure. The resulted device showed a PCE of 9.5%. 66 Zhang et al. incorporated Bi 3+ with varying concentration from 1 to 10 mol. % in CsPbI 3 to form CsPb 1−x BixI 3 compounds [lattice structure is shown in Fig. 9(b) (left)]. 142 They obtained CsPb 0.96 Bi 0.04 I 3 with 4 mol. % Bi 3+ and achieved an exceptional PCE of 13.21% via employing inorganic CuI as HTM in a FTO/c-TiO 2 /perovskite/CuI/Au structure [ Fig. 9(b) (right)], which is the highest efficiency of an all-inorganic triiodide PSC among all reported results.
Huang et al. fabricated HTL-free inorganic CsPbIBr 2 PSC with glass/FTO/c-TiO 2 /CsPbIBr 2 /Au structure by using dual source thermal evaporation and obtained a PCE of 4.7%. 28 Our group designed the FTO/NiOx/CsPbIBr 2 /MoOx/Au structure using the evaporated MoOx (4 nm) layer serving as a cathode buffer layer, which decreased the Schottky barrier, contact resistance, and interface trap-state density, thereby increasing the PCE to 5.52%, from the initial 1.3%. 143 Zhang et al. processed the CsPbIBr 2 film with light before thermal annealing; as a result, the PCE of device prepared by these novel methods reached a value up to 8.60%. 144 They had also undertaken an intermolecular exchange procedure by tuning the concentration of the CsI solution to acquire a full-coverage CsPbIBr 2 film featuring high crystallinity. Hence, CsPbIBr 2 -based all-inorganic PSC showed a PCE as high as 9.16%. 145 Soon after, they made a leap in PCE, increasing the efficiency to 10.71% by optimizing the energy alignment at the TiO 2 /perovskite interface. 146 153 As shown in Fig. 9(c), the conduction band minimum (CBM) of CsPbI 2 Br (4.16 eV), ZnO (4.2 eV), and C 60 (4.5 eV) is arranged in a gradient; thus, the electron extraction in the ZnO@C 60 bilayer is enhanced in comparison to the situation with an individual C 60 layer or ZnO layer. It is worth noting that a two-step temperature-control method was proposed in our film preparation procedure. The spin-coated CsPbI 2 Br thin film was actually heated to 42 ○ C to obtain the transition film, which was followed by thermal annealing at 160 ○ C for 10 min to obtain the desired dark-brown CsPbI 2 Br thin film with high orientation; the process is shown in Fig. 9(d). Similar two-step temperature-control methods have been widely used by other research groups. 18,107,121 In fact, we have previously adopted a two-step temperature-control method in which we prepared the CsPbIBr 2 solar cell, but the first step was operated at 30 ○ C. Similarly, we applied an optimized two-step temperaturecontrol method as a doping procedure of the perovskite host crystal, as follows. Different from the previous low temperature control procedures, the process was finalized by thermal radiation annealing {without bringing the substrate in contact with the hot plate [ Fig. 9(e)]}, allowing a perfect phase transition procedure under uniform thermal exposure, which is advantageous for the fabrication of large area devices (thus addressing the challenge in preparing large scale inorganic PSCs). 154,155 In this report, we demonstrated that the δ-phase perovskite can be inhibited by incorporating InCl 3 into the host perovskite lattice and thereby enhancing the phase stability. The evolution of the crystal structure and the space group of the CsPbBr 3 single crystal with the InCl 3 dopant are shown in Fig. 9(f). The enhanced stability was achieved by a strategy for structural reconstruction of the CsPbI 2 Br perovskite by means of In 3+ and Cl − codoping, which gave rise to a significant improvement in the overall spatial symmetry, with a closely packed atom arrangement due to the crystal structure transformation from the orthorhombic to cubic phase. The InCl 3 -based device displayed a maximum efficiency of 13.74%, which is, to our knowledge, among the best PCEs for all-inorganic PSCs [ Fig. 9(g)].

IV. STABILITY ISSUES
The stability issues of PSCs manifest roughly in two aspects: (1) inherent instability caused by physically and chemically unstable absorbers; perovskite materials are prone to react with water molecules in air that causes distortion of the corner-sharing octahedral framework ([PbI 6 ] 4− ) as well as thermal decomposition and light induced phase separation and (2) instability caused by instability of functional layers other than the absorber.

A. Absorber stability issues
Light-induced phase separation is common in organicinorganic perovskites. McGehee et al. is the first group to study the light stability of inorganic perovskites. A series of CsPbI 3−x Brx films were used for light stability measurements. The α-phase perovskites tend to transform to the insulating δ-phase perovskite at room temperature when x < 0.6, and the perovskites stay stable when 0.6 < x < 1.2 but phase segregation occurs and both I − rich and Br − rich domains appear under illumination when 1.2 < x < 3. 156 Zhao et al. also demonstrated that the CsPbI 2 Br perovskites have excellent stability under light exposure compared to MA-based perovskites. 43 They monitored a CsPbI 2 Br film and a MAPbBr 3 film by optical imaging during the poling process in ambient air at 25 ○ C. Degradation for the CsPbI 2 Br film was not detected either in the dark or under illumination (5 and  As discussed above, organic-inorganic perovskites tend to decompose under thermal environment, due to the volatility of organic A + cations. In order to have any impact in the market, commercial photovoltaic devices should be able to work effectively above 85 ○ C. Fortunately, inorganic materials always withstand these rather high temperatures. Cahen et al. compared hybrid organicinorganic MAPbBr 3 and inorganic CsPbBr 3 directly in terms of their thermal characteristics. The thermogravimetric analysis (TGA) results showed that CsPbBr 3 exhibits an obviously higher thermal stability than MAPbBr 3 [ Fig. 10(d)]. The temperature range in which CsPbBr 3 is thermally stable is up to 550 ○ C. It is worth noting that the decomposition arises from the loss of PbBr 2 rather than CsBr. 30 In 2016, the group of Snaith group demonstrated that the CsPbI 2 Br thin film with optimal crystallinity and absorbance could endure sintering temperature up to 400 ○ C. 27 To verify that the thermal stability of inorganic perovskite films is superior to that of organic-inorganic perovskites, they displayed the transformation of absorption intensity and XRD spectra of CsPbI 2 Br and MAPbI 2 Br films over time at 85 ○ C in air (20%-25% RH). The absorption intensity for CsPbI 2 Br films at 627 nm, i.e., the peak at the onset of absorption, remained approximately constant after 270 min of heating [ Fig. 10(e)]. As shown in Fig. 10(g), the XRD pattern almost did not show change and no new peaks occurred after heating, which indicated the absence of degradation of the CsPbI 2 Br films. In contrast, for MAPbI 2 Br, a continuous decrease in the perovskite absorbance at 670 nm during 270 min of heating was shown, as reproduced in Fig. 10(f). The XRD patterns showed perovskite peaks for MAPbI 2 Br before and after heating. In Fig. 10 Results indicate that CsPbI 2 Br shows wonderful phase stability compared with MAPbI 2 Br under identical conditions, namely, 85 ○ C and 20%-25% RH. However, the operational stability of CsPbBr 3 is proved to be better than MAPbBr 3 by measuring the photocurrent density as a function of time under 60%-70% RH. 30 The device stability under electron beam was further analyzed using electron beam-induced current (EBIC) analysis. Here, EBIC acted to generate electron-hole pairs, equivalent to the action of a light source. The electrons and holes are then separated in the junction area into free carriers, which were collected at the contacts. Clearly, there was almost no degradation in CsPbBr 3 , while MAPbBr 3 -based cells exhibited a decrease in collection efficiency.
In fact, the instability in a humid environment is a serious problem for inorganic perovskite materials. The stability of devices under ambient atmosphere of four types of inorganic perovskite cells (CsPbBr 3 , CsPbIBr 2 , CsPbI 2 Br, and CsPbI 3 ) was reported by Yuan et al., as shown in Fig. 11(a). The results explicitly showed that, during the test period of 250 h, both CsPbBr 3 and CsPbIBr 2 perovskites exhibit fairly good device stability, compared to CsPbI 2 Br

RESEARCH UPDATE
scitation.org/journal/apm and CsPbI 3 perovskites. Importantly, α-phase CsPbI 3 films immediately converted to the δ-phase when exposed to ambient conditions. Many groups have taken steps to address the problem that α-phase CsPbI 3 tends to transform to the insulating δ-phase perovskite at room temperature. HI was added in the perovskite precursor solution to generate lattice strain and form perovskite with small grains. 67 Controlling the electronic coupling of perovskite films with QDs of well-controlled size can increase storage time. 26,71 Reducing perovskite dimension can avoid the undesirable formation of the nonperovskite δ-phase. 77,79,81 Tetragonal β-phase and orthorhombic γ-phase perovskite films with lower surface free energy were fabricated to stabilize the devices. 17,[92][93][94][95][96] The best-performing γ-phase CsPbI 3 :Cl 0.03 solar cell reported by the group of Liu still maintained over 94% of its initial efficiency over a period of testing of 60 days under exposure to 20%-30% RH conditions [ Fig. 11(b)]. 17 Perovskite host doping gave rise to structural reconstruction and improved lattice stability. The best device based on CsPbI 3 -OTG 3 without any encapsulation still retained about 85% of its original PCE after being aged for 30 days in the dark at RH less than 10%. 157 Surface modification of functional groups improved stability. The PVP-modified cubic CsPbI 3 PSC that had a maximum efficiency of 10.74%, remained at 75% of initial PCE after storage for 500 h under 45%-55% RH conditions [ Fig. 11(c)]. 99 Although the PCE of CsPbI 3 has surpassed 17% and the perovskite phase can be stabilized in air after plentiful improvements, the RH of the operating environment still needs to be very low, and some of the materials cannot even be stored in inert atmosphere. If the RH of the operating environment is even slightly elevated, the devices are limited by their reduced RESEARCH UPDATE scitation.org/journal/apm with deliquescence and hygroscopic nature absorbs water in air and causes rapid decay of the HTL as well as the perovskite film, influencing the overall stability of the device. 158 The tBP has a relatively low boiling point, which easily causes evaporation under both device fabrication and storage conditions, 159 and it also forms coordinated complexes with PbI 2 [PbI 2 ⋅tBP], thus giving rise to chemical decomposition of perovskite. 160 Currently, most efforts focus on the stability of perovskite films but ignore the stability issues of other function layers. The selection of inorganic ETLs and HTLs is a particularly effective method to boost the stability of the devices, which clearly revealed the significance of research in all-inorganic PSCs. All-inorganic CsPb 0.97 Sm 0.03 Br 3 PSCs showed excellent stability either at a high RH of 80% or under continuous heating at 80 ○ C, as shown in Fig. 12(a). 60 The stability of this device is also dependent on the natural stability of tribromo-based perovskite. The all-inorganic CsPb 0.96 Bi 0.04 I 3 PSC with a FTO/c-TiO 2 /CsPb 0.96 Bi 0.04 I 3 /CuI/Au structure maintained 68% of its initial PCE value after 168 h under average RH of 55% at 25 ○ C. Photographs of the corresponding devices under exposure to air for 0, 74, and 168 h were shown in Fig. 12(b). 142 All-inorganic CsPbI 2 Br PSC based on a FTO/NiOx/CsPbI 2 Br/ZnO@C 60 /Au structure maintained 80% of its initial PCE after aging under continuous heating at 85 ○ C for 360 h [ Fig. 12(c)]. 153 For the all-inorganic PSCs with InCl 3doping, we conducted stability tests under different moisture and temperature conditions. The devices maintained their initial PCEs in RH = 30%-40% air at room temperature for over 80 h and declined 20% relative to the initial efficiency after 100 h at 60 ○ C with RH ≈ 50% [ Fig. 12(d)]. 154 Such stability under both high temperature and moisture is rare in PSCs based on inorganic materials due to the limitations of organic additives and organic materials.
The metal (such Au and Ag) electrode has been proved to easily react with the halide ions in perovskites, resulting in a decrease in device efficiency. 161,162 On the contrary, a carbon electrode has exhibited more impressive promise, including a low-cost, simple process, suitable work function, 163 high resistance to water, and inertness to halide ions in perovskite, which all help to enhance device stability. 164 The PCE of CsPbBr3/carbon-based all-inorganic PSCs exhibited a remarkable stability during a long testing period of 840 h under a high temperature of 100 ○ C. 29 However, compared to devices with a metal electrode, the efficiency of perovskite devices based on the carbon electrode needs to be further improved.
These data suggest that a combination of inorganic perovskite absorbers and inorganic functional layers as well as a carbon electrode is promising for achieving high thermal stability.

V. CONCLUSION AND PERSPECTIVE
Solar cells based on inorganic perovskite materials have achieved much progress, both in PCE and stability, in a short time. Encouragingly, the PCE of CsPbI 3 and CsPbI 2 Br PSCs has been improved to 18.4% and 16.2%, respectively, which makes them the two most promising candidates for commercial applications (a PCE of 15% is generally believed to be the threshold value for application). Nevertheless, there is still considerable room to improve the efficiency for solar cells based on these inorganic perovskite materials due to their wider bandgaps. To decrease the bandgap, the incorporation of metal ions with a small radius at the lattice sites of Pb 2+ has been pursued. Unfortunately, when the amount of metal ions is small, this leads to a negligible adjustment to the bandgap, while a large number of substitutions reduces the bandgap at the cost of efficiency, due to the known tendency that Sn 2+ /Ge 2+ is easily oxidized to Sn 4+ /Ge 4+ . A 2 SnX 6 , perovskite derived material, addresses the problem that Sn 2+ is easily oxidized; however, this is only the first step in a long march and large effort needs to be dedicated to further research.
Interestingly, inorganic perovskite materials present advantages in terms of light and thermal stability. Especially, CsPbBr 3 possesses excellent moisture stability. The promising CsPbI 3 and CsPbI 2 Br PSCs still have deficient stability under moist conditions; thus, crucial contributions are still needed to achieve long-term stability.
The overall stability of the devices depends on all absorber and functional layers. At present, most of the solar cells based on inorganic perovskite materials take the traditional n-i-p structure as a framework and inevitably make use of organic materials with volatile and hygroscopic nature. Replacing the undesired organic layers with inorganic layers and forming all-inorganic PSCs is a trend to fundamentally tackle the disturbing instability issues. Nevertheless, most of the HTLs that are suitable for applications in traditional structures are organic compounds. In this regard, the inverted p-i-n all-inorganic PSC structures, which have the potential to be mass produced, provide a possible path toward stable PSCs. Although there are only few reports available on all-inorganic PSCs with the p-i-n structure, it would not be surprising if they are going to be widely reported in the next few years. The main difference between p-i-n and n-i-p structured devices lies in the Voc, and the voltage deficit in the former structure has always been higher than that of the latter structure, while it is unknown what this is due to. Ideally, investigations on how to diminish the Voc loss in p-i-n structured all-inorganic PSCs should prominently be put on the agenda.