Lead‐Free Hybrid Perovskite Absorbers for Viable Application: Can We Eat the Cake and Have It too?

Abstract Many years since the booming of research on perovskite solar cells (PSCs), the hybrid perovskite materials developed for photovoltaic application form three main categories since 2009: (i) high‐performance unstable lead‐containing perovskites, (ii) low‐performance lead‐free perovskites, and (iii) moderate performance and stable lead‐containing perovskites. The search for alternative materials to replace lead leads to the second group of perovskite materials. To date, a number of these compounds have been synthesized and applied in photovoltaic devices. Here, lead‐free hybrid light absorbers used in PV devices are focused and their recent developments in related solar cell applications are reviewed comprehensively. In the first part, group 14 metals (Sn and Ge)‐based perovskites are introduced with more emphasis on the optimization of Sn‐based PSCs. Then concerns on halide hybrids of group 15 metals (Bi and Sb) are raised, which are mainly perovskite derivatives. At the same time, transition metal Cu‐based perovskites are also referred. In the end, an outlook is given on the design strategy of lead‐free halide hybrid absorbers for photovoltaic applications. It is believed that this timely review can represent our unique view of the field and shed some light on the direction of development of such promising materials.

approach to get sustainable clean energy safely. As one of the third-generation PV technologies, hybrid halide perovskite solar cells (PSCs) emerged since Miyasaka and co-workers [1] incorporated MAPbX 3 (X = I, Br) as sensitizers into dye-sensitized solar cells (DSSCs), achieving a power conversion efficiency (PCE) of 3.8% in 2009. After numerous research endeavors in the past eight years, the PCEs of PSCs rapidly improved to 22.1%. [2] Originally, hybrid halide perovskites with the general formula of ABX 3 are structural analogs of the natural mineral CaTiO 3 , while A is a monovalent, organic or alkali metal cation, M is a divalent p-block metal (typically Pb, Sn, and Ge), and X is a halide anion. [3] Depending on the demand of the researchers, hybrid halide perovskites and their derivatives for the photovoltaic application can be classified into three main categories since 2009 (Figure 1).
Aforementioned 3D lead halide perovskites realized the highest-performing solution-processed solar cell on record, rivaling commercial crystalline silicon solar cells in efficiency. [26] However, the toxicity issue of the lead urged some researchers to seek alternatives to lead-based perovskites. We classify these alternatives as the second group of perovskites, which features less toxic lead-free hybrid halide light absorbers. Lead-free hybrid halide light absorbers mainly include group 14 metals like tin (Sn) and germanium (Ge), group 15 posttransition metals like bismuth (Bi) and antimony (Sb), and transition metal copper (Cu) as the metal cations. [27] In this case, a variety of crystallographic polymorphs appeared: Sn-and Gebased compounds with 3D perovskite framework; Bi-and Sbbased "pseudoperovskite" without corner-shared MX 6 octahedra structure; Cu-based typical 2D layered perovskites. As light absorbers used in solar cells, Sn-based perovskites achieved the highest efficiency so far of 8.12% among all lead-free hybrid halide compounds. [28] In the group of high-efficiency lead halide perovskites, there is another problem of insufficient long-term stability for the application of the devices. One solution to this problem is mixing 3D perovskites with 2D perovskites, [29][30][31][32] which can be categorized as the third group of perovskites with the mission of realizing high efficiency and stability simultaneously. The mixed dimensional (MD) perovskites have a general chemical formula [30] of (A) 2 (CH 3 NH 3 ) n−1 MX 3n+1 (n is an integer), where A is a primary aliphatic or aromatic alkylammonium cation, M is a divalent metal, and X is a halide anion. In the MD perovskites, the large organic cations (A) defragment the 3D structure and isolate certain number (n) of inorganic perovskite layers of corner-sharing [MX 6 ] 4− octahedrons. [33] This configuration was found to prevent moisture from attacking the perovskite and therefore improve the stability of perovskite film. Additionally, the wide variety of "A and n" brings MD perovskites abundant tunability and flexibility to control the physical properties, as well as balanced stability versus optoelectronic performance of corresponding devices. So far, the quasi-2D PEA 2 (CH 3 NH 3 ) n−1 Pb n I 3n+1 [32] and 2D perovskite (BA) 2 (MA) 3 P b 4 I 13 [34] displayed good efficiency over 12% through optimizing stoichiometry of materials and showed much improved stability than intrinsic 3D perovskites.
The three groups of hybrid perovskites attracted attention to different extents. In this series review, we focus on the second group of perovskites or lead-free hybrid light absorbers used in photovoltaic devices. Recent development in structures, optoelectronic properties, and the related solar cell applications of these types of hybrid light absorbers are summarized. In the first part, group 14 metals (Sn and Ge)-based perovskites are introduced with more emphasis on the optimization of Sn-based PSCs. Then we put concerns on halide hybrids of group 15 metals (Bi and Sb), which are mainly perovskite derivatives. At the same time, transition metal Cu-based perovskites are also referred. In the end, we give an outlook on the design strategy of lead-free halide hybrid absorbers for photovoltaic applications (Figure 2). An almost complete summary of the state-of-the-art development of the lead-free halide hybrid absorbers in PV devices is listed in Table 1.

Lead-Free Halide Hybrid Perovskite and Related Absorbers
Although we have achieved high efficiency beyond 22% based on lead halide perovskites, which is comparable to commercial crystalline silicon solar cells, the existence of Pb is an urgent problem against the final application of PSCs. The toxicity of lead is documented to disturb the functioning of the blood, kidneys, liver, testes, brain, and nervous system. [124][125][126] The toxicity of lead is due, in general, to its binding affinity to thiol and cellular phosphate groups of numerous enzymes, proteins, and cell membranes. [127] Lead is toxic to the central nervous system, especially in children. [128] Thus, there come the questions that: can we eat the cake and have it too? Meaning can we have highefficiency PSCs without being poisoned? The urge to explore authentic high-efficiency lead-free metal halide absorbers led to an abundance of works which will be shown in the following sections. and α-phase structure. [132,136,137] Another unique point is that due to the symmetry reduction of the 3D [SnI 3 ] − framework with larger cations group like FA, [132] FASnI 3 has a larger band gap (1.41 eV) value than that of MASnI 3 . [59] In 2015, Koh et al. first took FASnI 3 as a light absorber in solar cell applications and realized a high short-circuit current density (J SC ) of 24.45 mA cm −2 and PCE of merely 2.1%. [59] It was reported that FASnI 3 is more stable than MASnI 3 due to the reduced extent of Sn oxidation . [133,138] Accordingly, FASnI 3 -based PSCs exhibited much better reproducibility as compared to MASnI 3 -based devices. [138] In 2016, Liao et al. [56] reported a PCE of 6.22% based on FASnI 3 PSCs with high reproducibility, which is one of the best efficiencies among Sn-based PSCs (Figure 3d,e). Very recently, Shi et al. [139] studied the phenomenon theoretically and found that the antibonding coupling between Sn-s and I-p is weaker in FASnI 3 than in MASnI 3 due to the larger ionic size of FA, leading to higher formation energies of Sn vacancies in FASnI 3 . Subsequently, the conductivity of FASnI 3 can be tuned from p-type to intrinsic by varying the growth conditions of the perovskite semiconductor. In contrast, MASnI 3 shows unipolar high p-type conductivity independent of the growth conditions. Ion mixing is also one important approach for Sn perovskites. Ferrara et al. [61] reported the first mixed A-cation compositions of tin perovskites FA 1−x MA x SnBr 3 , with cubic structures and the band gaps ranging from 2.4 eV (x = 0) to ≈1.92 eV (x = 0.82). However, no device performance was reported. Very recently, Zhao et al. [28] reported another mixed-organic-cation perovskite absorber (FA) x (MA) 1

CsSnI 3
Compared to the organic-inorganic hybrid perovskite materials, the all-inorganic halide perovskites exhibited much higher thermal stability. [16,70,140] Replacing the organic cations (MA + or FA + ) with the inorganic Cs + , CsSnI 3 perovskite shows a melting point of 435 °C, [72,141] suggesting superior intrinsic thermal stability. In contrast, hybrid perovskites MASnI 3 and FASnI 3 start to decompose at ≈200 °C. CsSnI 3 adopts a black orthorhombic perovskite phase structure, possesses a direct band gap of 1.3 eV, and has a high hole mobility of ≈585 cm 2 V −1 s −1 at room temperature (Figure 5a). In 2012, due to its high hole mobility, it was first taken as a hole transport material (HTM) in DSSCs. [72] Very recently, the further potential of CsSnI 3 for solar cell applications was uncovered by Wu et al. [142] They reported that the melt-synthesized CsSnI 3 ingots contain high-quality large single crystal (SC) grains, which bear superior properties like high bulk carrier lifetimes (>6.6 ns), low doping concentrations of ≈4.5 × 10 17 cm −3 , and long minority-carrier diffusion lengths approaching 1 µm. In this regard, they predicted a PCE of ≈23% for optimized CsSnI 3 SC solar cells. As a light absorber, CsSnI 3 was firstly used in a Schottky contact type PSC with a simple layer structure of ITO/ CsSnI 3 /Au/Ti in 2012, showing a PCE of 0.9%. [71] Four years later, Marshall et al. [66] designed an HTM-free CsSnI 3 PSC with exceptionally high fill factor up to 0.69, and a PCE up to 3.56%. In 2017, Song et al. [65] assembled CsSnI 3 -based PSCs with reducing atmosphere-assisted dispersible additive. The new technique produced a PCE of up to 4.81%, which is the highest efficiency among all CsSnI 3 PSCs so far ( Figure 5a). Meanwhile, benefiting from the superior thermal stability of CsSnX 3 perovskite, Li et al. [73] fabricated an all-inorganic CsSnIBr 2 mesoscopic PSC with thermal stability up to 200 °C, achieving an average PCE of 3.0% and longterm stability without efficiency-loss over 77 d (Figure 5c).

Cs 2 SnI 6
The Sn II in CsSnI 3 and M(F)ASnI 3 has the tendency to be oxidized to Sn IV leading to the spontaneous change to Cs 2 SnI 6 and M(F)A 2 SnI 6 respectively in ambient atmosphere. Cs 2 SnI 6 with a molecular salt structure is not bona fide perovskite but crystallizes into a face-centered cubic 3 Pm m space group and the lattice parameter of which is 11.63 Å [132,144] (Figure 5d). The direct bandgap (E g ) of Cs 2 SnI 6 is ≈1.3-1.6 eV [76,79,132] depending on the preparation methods at room temperature with absorption coefficient over 10 5 cm −1 from 1.7 eV. Computational work based on Cs 2 SnI 6 [145] shows that the iodine vacancies and interstitial Sn are the dominant defects that give intrinsic n-type behavior, unlike the p-type behavior in CsSnI 3 . The carrier concentration (electrons) of Cs 2 SnI 6 is on the order of ≈1 × 10 14 cm −3 measured by Hall effect measurements at room temperature, with a high electron mobility of 310 cm 2 V −1 s −1 . Moreover, due to the full oxidation state of Sn 4+ , Cs 2 SnI 6 exhibits better stability in the air with moisture than CsSnI 3 . In 2014, Lee et al. used Cs 2 SnI 6 as a hole transporter in solid-state DSSCs and fabricated devices in the air with an efficiency close to 8%. [78] The study of Cs 2 SnI 6 as    Reproduced with permission. [133] b) The spectra of MASnI 3 and FASnI 3 prepared with the solution method as compared with other perovskites. Reproduced with permission. [132] Copyright 2013, ACS. c) J-V curve of MASnI 3 on TiO 2 and bandgap of the material determined using the Tauc plot. Reproduced with permission. [46] Copyright 2014, RSC. d) Cross-sectional SEM image of the entire device with FASnI 3 and 10 mol% SnF 2 additive; e) J-V characteristics of FASnI 3 -based devices under 100 mW cm −2 AM1.5G illumination under reverse and forward voltage scans. Reproduced with permission. [56] a light absorber in the solar cells is rare. Until 2016, Cs 2 SnI 6 was first studied by Qiu et al. [76,77] as a light harvester, demonstrating a PCE of ≈1%. Very recently, Lee et al. [80] further improved the PCE up to 1.47%. They designed a series of compounds with the general formula of Cs 2 SnI 6−x Br x . With the increase of Br composition (x), the bandgaps can be tuned from ≈1.3 to ≈2.9 eV and the color of the films was changing from dark brown to brown/red, then to light yellow. The Cs 2 SnI 6−x Br x films were fabricated with a two-step solution process: the crystal structure of CsI was optimized in Step-1 by postannealing at 300 °C for 30 min after electrospraying deposition and in Step-2 the CsI film was reacted with a SnI 4 solution at 110 °C for 20 min (Figure 5e,f). After that, a stoichiometric, smooth, uniform, and thick active layer was obtained. The freshly made films were constructed into the typical "sandwich" type device structure of FTO/bl-TiO 2 /2wt% Sn-TiO 2 /Cs 2 SnI 6−x Br x /Cs 2 SnI 6 HTM/largeeffective-surface-area polyaromatic hydrocarbon (LPAH)/FTO. The best-achieved efficiency was ≈2.03% when x = 2. It is worth noting that their device fabrication process was carried out in the air and did not use any additive to protect the active material. The Cs 2 SnI 6−x Br x films and corresponding devices showed excellent stability in the air for 50 d (see Figure 5f). Thus, being a molecular salt (0D), Cs 2 SnI 6 has the bandgap similar to that of the 3D perovskites (CsSnI 3 and MAPbI 3 ), high absorption coefficient, and high carrier mobility. Coupled with its intrinsic ambient stability, such Sn-perovskite variants can be explored Reproduced with permission. [65] Copyright 2017, ACS. c) Bandgap variation with Br concentration. Reproduced with permission. [70] Copyright 2015, ACS. d) Distorted 3D structure of Cs 2 SnI 6 at room temperature. Reproduced with permission. [78] Copyright 2014, ACS. e) J-V curves of a series of cells with different composition of Cs 2 SnI 6−x Br x . The inset shows the IPCE values.
f) The stability curves for 50 d are shown here for Cs 2 SnI 6 (black), and the Cs 2 SnBr 2 I 4 (green)-based solar cells. Reproduced with permission. [80] Copyright 2017, RSC. Reproduced with permission. [28] with more effort in the future for achieving more efficient and stable Sn-based PSCs.

Germanium-Based Absorbers
Germanium, another group IV metal, with ns 2 electronic configuration, has the same valent state with the lead. Due to the 4s lone pairs of Ge is more active than the Pb 6s lone pair, Ge 2+ is easier to be oxidized leading to metallic conductivity in the hybrid materials and short-circuit behavior in the photovoltaic devices similar to the case of Sn-based perovskite. [40,135,141] However, germanium has demonstrated much less toxicity compared to lead [146] and is expected to be a promising candidate in the search for Pb-free perovskite materials. To prove the concept, Sun et al. [147] investigated the structural and electronic properties of MAGeX 3 (X = Cl, Br, I) by density functional theory (DFT) methods and showed that MAGeI 3 is a good absorber for applications in PSCs. Based on DFT calculation, Ming et al. [148] also proposed that CsGeI 3 might be a good HTM in solar cells. Krishnamoorthy [40] studied the solid structure of three AGeI 3 (A = Cs, MA, or FA) halide perovskites and revealed trigonal phase (with R3m space group symmetry) in contrast to Pb and Sn-based perovskites with a tetragonal phase (I4/mcm) [147] at room temperature. All compounds are remarkably stable up to 150 °C and show no phase transition in the range of device working temperatures (r.t. to 150 °C). With increasing size of the A + cation, the band gaps of AGeI 3 are 1.63, 2.0, and 2.35 eV for CsGeI 3 , MAGeI 3 , and FAGeI 3 , respectively (Figure 6a). The values of the valence bands of CsGeI 3 , MAGeI 3, and FAGeI 3 perovskites measured by the PESA are −5.10, −5.2, and −5.5 eV, respectively ( Figure 6b). However, unfortunately, the solar cells based on AGeI 3 were fabricated with a mesoscopic structure, achieving 0.2% PCE for MAGeI 3 , 0.11% PCE for CsGeI 3 and no photocurrent for FAGeI 3 (Figure 6c). The low performance of AGeI 3 -based solar cells was attributed to the low concentration of precursor solutions, poor quality of perovskite films, and the oxidation sensitivity of the materials. Recently, theorists studied mixed tin and germanium perovskites and predicted that RbSn 0.5 Ge 0.5 I 3 possesses not only a direct bandgap within the optimal range of 0.9-1.6 eV but also a desirable optical absorption spectrum that is comparable to those of the state-of-the-art MAPbI 3 perovskites. It has favorable effective masses for high carrier mobility as well as a greater resistance to water penetration than the prototypical inorganic-organic lead-containing halide perovskite. [149]

Devices Engineering Efforts Toward High Efficiency and Stable Sn 2+ -Based PSCs
Thanks to the great endeavor of the researchers, Sn-based PSCs have obtained relative stable PCE of ≈8%. [28] The improvement in performance of PSCs is not only due to the material itself but also the evolution in device engineering. In this small section, we will introduce various devices preparation approaches towards efficient and stable Sn 2+ -based PSCs over the recent years.

Efforts to Suppress Sn Vacancy Defects and Improve
Oxidation Stability: One of the major problems of Sn-based PSCs is their poor stability originating from the facile formation of Sn vacancy associated with oxidation of Sn 2+ to Sn 4+ when exposed in air, [46,141,150] which leads to poor reproducibility and deteriorates the devices rapidly. [46,51] For example, Xu et al. [151] studied the defect properties of CsSnI 3 perovskite Adv. Sci. 2018, 5, 1700331 Figure 6. a) Optical absorption spectrum of CsGeI 3 , MAGeI 3, and FAGeI 3 , in comparison with CsSnI 3 ; b) Schematic energy level diagram of CsGeI 3 , MAGeI 3, and FAGeI 3 ; c) J-V curves of photovoltaic devices fabricated with different germanium halide perovskites. Reproduced with permission. [40] Copyright 2015, RSC. and depicted the influence of defects and synthesis conditions on the photovoltaic performance. They found that due to the strong Sn 4s-I 5p antibonding coupling, Sn vacancies have very low formation energies in CsSnI 3 , leading to a very high concentration of Sn vacancies and therefore high p-type conductivity regardless of the growth conditions. To solve the problem, reductive or sacrificial additives like Sn(II) halide salt etc., were added to cure the problem.

SnF 2
Before tin halide perovskite served as a light-absorber in PSCs, Chung et al. [72] demonstrated that when CsSnI 3 with SnF 2 additive (though the formulation of CsSnI 2.95 F 0.05 is not correct [152] ) was used as HTM in DSSC, enhancements of 29% and 21% in J SC and η were achieved, respectively. The magic effect of SnF 2 attracted researcher to investigate its working mechanism. In 2014, Kumar et al. [69] demonstrated for the first time that the introduction of SnF 2 into CsSnI 3 can reduce Sn vacancies and background carrier densities. Therefore, high J SC of more than 22 mA cm −2 and a PCE of 2.02% were achieved in contrast to the nonfunctioning devices without SnF 2 . One year later, the same group [59] further investigated the doping effects of SnF 2 in FASnI 3 . The X-ray photoelectron spectroscopy (XPS) data show that only Sn 2+ but no Sn 4+ appears in the FASnI 3 : 20% SnF 2 sample, indicating that no oxidation side reaction happened inside the doped sample (Figure 7a-d). Moreover, they claimed that the addition of SnF 2 improved the environmental stability, as they found that the color of the FASnI 3 films with SnF 2 remained yellow when left overnight, but change into orange without SnF 2 (Figure 7e). The resulting photovoltaic devices gave maximum PCE of 2.10%, with high J SC of 24.45 mA cm −2 . Based on the same hypothesis, SnF 2 dopant was also introduced to reduce Sn vacancies for efficient and stable CsSnBr 3 based PSCs. [74,75] The author has concluded that the SnF 2 cannot only improve interfacial energy alignment but also increase stability to electron beam damage. Nowadays, most Sn-based perovskites films prepared by solution deposition need additional SnF 2 to get good performance, [47,48,57,58,73] including the recent (FA) 0.75 (MA) 0.25 SnI 3 based PSCs with the highest PCE of 8.12%. [56] Additionally, Ma et al. [134] claimed that SnF 2 as the inhibitor of Sn 2+ oxidation in the MASnI 3 film could increase the fluorescence lifetime up to 10 times and give longer carrier diffusion lengths >500 nm, compared with pristine MASnI 3 films.

SnX 2 (X = Cl, Br, I)
Apart from the tin fluoride additives, excess tin halide salts like SnI 2 were also believed to improve device stability towards air oxidation. In 2015, Marshall et al. [68] prepared the CsSnI 3 films with an excess of SnI 2 during CsSnI 3 synthesis from CsI and SnI 2 . They believed that the excess of SnI 2 occupies some of the space between adjacent CsSnI 3 crystals, hindering the ingress of the oxygen and water so that the barrier of the transformation of CsSnI 3 into Cs 2 CsI 6 at the site of a Sn vacancy is increased. The fabricated devices have inverted planar structure of ITO/CuI/CsSnI 3 (with or without the excess of SnI 2 )/C 60 / BCP/Al. The devices with the excess amount of SnI 2 exhibit up to 30% increase in J SC and V OC as compared to the control devices. The devices' stability is also enhanced in that the PCE of devices with 10 mol% excess SnI 2 exhibited only 10% reduction after 10 d period of storage, contrary to the 70% loss with control devices. In 2016, they further demonstrated that excess SnI 2 is beneficial to improve both efficiency and stability of CsSnI 3 -based PSCs. [66] Therefore, their results suggest that the excess amount of SnI 2 in the precursor is also an effective  strategy to improve the performance of CsSnI 3 -based PSCs. Similarly, an excess of SnBr 2 [75] was also found to reduce the density of defect states or Sn vacancies in all-thermal-vapor deposited CsSnBr 3 PSCs by preventing the oxidation of Sn 2+ to Sn 4+ in ambient air. The resulted devices obtained a higher V OC of 0.41 V than the previously reported 0.19 V. [23] It was not until 2016 that a systematic study was made about the effect of different Tin halide salts in resolving the problem of tin vacancies in tin halide perovskites [66] (Figure 8a-f). Focusing on the low fill factor (FF) problem of Sn-based PSCs, Marshall et al. evaluated SnF 2 , SnCl 2 and SnBr 2 additives in CsSnI 3 -based PSCs. Simplified device architecture (ITO/CsSnI 3 /PC 61 BM/BCP/Al) was fabricated without HTM and the performance of SnCl 2 doped CsSnI 3 PSC devices was tested without encapsulation in ambient air at a humidity of ≈25% under constant 1 sun simulated solar illumination. They showed that the champion stability was exhibited only by devices with a tin halide additive: 11 h with no additive; 16 h with 10 mol% SnCl 2 ; and 22 h for SnI 2 (Figure 8g,h). Moreover, SnCl 2 as additive offers the advantage of the highest η (3.56%, due to reduced sensitivity of device parameters to pin holes while SnBr 2 and SnF 2 gave η ≤ 0.4%. For the same reason, SnCl 2 -doped CsSnI 3 also showed FF up to 0.69, which is the highest FF among reported tin halide PSCs. They attributed SnCl 2 being the best of the tin halide additives to the interplay of three factors. First, the added SnCl 2 is distributed toward the surface of the crystallites to form a tin-rich top layer for the perovskites. Second, the greater covalence offers SnCl 2 better solubility in common solvents (including DMF) than SnF 2 . Last, tin chloride diffusion into fullerene (ETL) is easier than other halide analogs. Therefore, SnCl 2 can be used as a promising additive to improve the performance of CsSnI 3 PSCs and probably any other Sn-based PSCs. Currently, only one report on SnCl 2 modified FASnI 3 as an inhibitor of Sn 4+ , but poor PCE was achieved. [153] SnO and Sn(OH) 2 Except for the tin halide additives, the presence of SnO and Sn(OH) 2 [53] in the MASnI 3 film was also found to be beneficial to reduce hole carrier concentration, leading to an improved air stability of the Sn-based perovskite devices. This finding suggests that not only the commonly used tin halide additives but also other divalent Sn compounds could serve as Sn vacancy suppressors assisting the realization of efficient and stable tinbased PSCs.

SnX 2 /Organic Reduce Agents
The primary cause of the instability of tin-based PSCs is the oxidation of the Sn(II) to Sn(IV). Therefore, it is reasonable to add reductive agents besides divalent tin additives, to further suppress Sn vacancy for more stable lead-free PSCs. As an example, hypophosphorous acid (HPA) [73] (Figure 9a) was introduced into fabricating all-inorganic CsSnIBr 2 -based mesoscopic PSCs to reduce the concentration of Sn vacancies, achieving a higher PCE of 3.0% than 1.7% of without HPA addition. Moreover, Song et al. [41] introduced hydrazine vapor (reducing vapor atmosphere) in the presence of SnF 2 additive into fabrication process of Sn-based PSCs (Figure 9b). The Adv. Sci. 2018, 5, 1700331 Figure 8. SEM images of CsSnI 3 films on ITO glass prepared with different tin halide additives. SEM images with no a) tin halide additive, b) 10 mol% added SnI 2 , c) 10 mol% added SnBr 2 , d) 10 mol% added SnF 2 , and e) 10 mol% added SnCl 2 . f) Schematic diagram of the proposed film structure in case e: CsSnI 3 crystallites capped with a thin SnCl 2 layer. Reproduced with permission. [66] Copyright 2016, Nature Publishing Group. g) The effect of constant illumination on PCE of devices (100 mW cm −2 of AM 1.5) for 245 min. Reproduced with permission. [68] Copyright 2015, RSC. h) J-V characteristics of different CsSnI 3 devices after 60 d storage under nitrogen. Reproduced with permission. [66] Copyright 2016, Nature Publishing Group. results showed that the additional hydrazine vapor process led to significantly suppressed carrier recombination with more than 20% reduction of Sn 4+ /Sn 2+ ratios. And by tuning amounts of hydrazine vapor properly, the PCEs of MASnI 3 and CsSnI 3 devices were improved from an average of ≈0.02% to 3.40% and ≈0.16% to 1.50%, respectively. Four months later, the same group reported [65] more efficient CsSnI 3 cells with PCE up to 4.81% using the same hydrazine vapor treatment with an excess of SnI 2 . In this research, they also studied the optimum ratios of monovalence cations (MA + , FA + , or Cs + ) to SnI 2 in the MASnI 3 , FASnI 3 and CsSnI 3 perovskite materials, which were 0.4-0.6, 0.6-0.8, and 0.4, respectively. Furthermore, they prepared ASnI 3 devices in a pure N 2 atmosphere for comparison. The results showed that only FASnI 3 could work with V OC > 0.15 V, while the other two devices behaved short-circuit even at the optimum AI/SnI 2 (A: MA + , FA + , or Cs + ) ratio that worked in a weak hydrazine vapor atmosphere. This result showed the importance of using reducing agents like hydrazine to suppress SnI 2 from forming Sn 4+ and ensure Sn 2+ -rich environment to compensate the Sn 2+ vacancies for obtaining efficient devices. On the other hand, the phenomenon that only FASnI 3 -based devices could survive without additional hydrazine consists with aforementioned results [133,138] where FASnI 3 displayed alleviated self-doping effect possibly due to the competitive formation the hydrogen bonding between H 2 O and FA + (Figure 9c-e).

5-Ammonium Valeric Acid Iodide and Ascorbic Acid
In the end, beyond above-mentioned Sn(II) containing panaceas, Hoshi et al. [154] claimed that the addition of 5-AVAI in the preparation process could significantly improve the oxidation stability of the MASnI 3 films in the air, due possibly to the formation of (5-AVAI) x (CH 3 NH 3 ) 1−x PbI 3 . In addition, ascorbic acid (AA), [155] as a common antioxidant, was also introduced as an effective additive to retard the oxidation of Sn-containing precursor solutions for making Pb/Sn mixed perovskites. It is of great necessity to examine these effects on the pure Sn-based perovskites.

Methods to Control the Morphology of the Sn-Based
Perovskite Layers: Apart from the divalent tin additives, the film quality of the fabricated perovskite layer is another parameter that controls the final performance of the tin-based PSCs. The film quality mainly refers to the morphology of the perovskite layer, such as homogeneity and coverage, which are strongly influenced by the crystallization process of perovskite film. The conventional one-step film deposition method of tin perovskites often engenders randomly oriented film growth accompanied by forming large crystals platelets and poor surface coverage with micron-sized pinholes. [46] The flawed films further led to the poorly performing devices. [69] Moreover, due to the reaction kinetics between organic/inorganic halide and tin halide salts is faster than its lead analogs, [19,46,156] the control of tin perovskite crystallization during the deposition process is more challenging. Therefore, developing novel preparation methods to achieve high-quality tin perovskite film is of great importance to boost both efficiency and stability of Snbased PSCs. In lead-based perovskites, the methods for morphology engineering include addition of additives, [157][158][159] solvent engineering, [160,161] vacuum engineering, [162,163] thermal annealing, [164] self-healing, [165] vapor deposition, [166] and so on. Herein, the reported methods used for preparing high-quality tin perovskite layers are introduced.

Additives for the Morphology Control
Additives such as methyl ammonium chloride (MACl), [157] 1,8-diiodooctane (DIO), [158] and butyl phosphonic acid  4-ammonium chloride (4-ABPACl) [159] have been successfully used to obtain high-quality perovskite films for high-efficiency lead-based PSCs. In the case of tin perovskite, the popularly used SnF 2 additive for the elimination of the Sn vacancy could result in poor film morphology and bad device performance due to the phase separation of SnF 2 within the perovskite film. [6,135] In this regard, Lee et al. [58] introduced pyrazine into FASnI 3 perovskite precursors in conjunction with SnF 2 in the form of the SnF 2 -pyrazine complex (Figure 10a,b). The complex is believed to assist a uniform distribution SnF 2 in the perovskite film, thereby substantially improving the morphology of FASnI 3 perovskite. Finally, the resulted FASnI 3 PSCs achieved a maximum PCE of 4.8% with high reproducibility (Figure 10c). Besides being a reducing agent, hypophosphorous acid (HPA) can also act as a morphology controller [73] in the fabrication process of all-inorganic CsSnIBr 2 mesoscopic PSCs. HPA has the PO bond that could strongly coordinate with Sn 2+ , producing HPA-CsSnIBr 2 clusters via SnOPOSn coordination bonds. [167] The formed clusters in precursor solution promoted the growth of perovskite crystals and expelled the redundant SnF 2 to the surface of the film. Due to the suppressed SnF 2 phase separation in the CsSnIBr 2 thin films, the highest reported PCE of ≈3.2% for the all-inorganic Sn-based PSCs was achieved. Likewise, hydrazine [41,65] was used not only as reducing agent to reduce the oxidization of Sn 2+ to Sn 4+ , but also as modifier asserted by Song et al. [41] to achieve better film morphologies for enhanced device efficiencies. Unfortunately, there was no explanation of why hydrazine could do a good job in ameliorating perovskite film morphologies. Moreover, a hydrazine atmosphere can help the disperse of SnI 2 [65] Adv. Sci. 2018, 5, 1700331 Reproduced with permission. [58] Copyright 2016, ACS. d-i) SEM images of the CsSnI 3 perovskite films grown with various CsI/SnI 2 molar ratios, together with a neat SnI 2 film. Reproduced with permission. [65] Copyright 2017, ACS. Top surface SEM images of j) pristine 2D Sn 3 I 10 film, k) (Sn 3 I 10 + SnF 2 ) film, and l) (Sn 3 I 10 + SnF 2 + TEP) film. Cross-sectional SEM images of m) (Sn 3 I 10 + SnF 2 ) film and n) (Sn 3 I 10 + SnF 2 + TEP) film. Reproduced with permission. [63] Copyright 2017, ACS.
to improve the quality of perovskite film, as indicated by no observable agglomeration of SnI 2 from the SEM/EDS characterization (Figure 10d-i). The corresponding device displayed the best PCE (4.81%) of CsSnI 3 -based PSCs, which is in the meantime much higher than the SnF 2 [41] modified devices (1.83%). Additionally, triethyl phosphine (TEP) [63] as a soft Lewis base can form intermediate complexes with Sn 2+ species via weak coordinating interaction, which could slow down the perovskite crystallization process and improve the film morphology as well as the device performance. The devices with TEP showed increased FF in average (from 42.0% to 53.7%), and an average PCE from 1.15% to 1.75%, with the champion device reaching ≈2% efficiency (Figure j-n).

Solvent Engineering
Solvent engineering technique has been proved many times as the most effective method for preparing high-quality lead perovskite films to achieve high-performance PSCs. [160,161,[168][169][170] The work on solvent engineering of Sn-based perovskite was first reported by Hao et al., [47] where they investigated the solvent effects on the crystallization of the MASnI 3 perovskite films. They found that highly uniform, pinhole-free perovskite films can be obtained by using a dimethyl sulfoxide (DMSO) as solvents in perovskite precursor. And the transitional SnI 2 ·3DMSO intermediate phase was very important in achieving a high-quality perovskite film. The heterojunction depleted hole-transporting layer-free solar cells based on mesoporous TiO 2 showed a high photocurrent up to 21 mA cm −2 and a PCE of 3.15%. Furthermore, Lee et al. [58] used mixed solvents of N,N-dimethylformamide (DMF) and DMSO in FASnI 3 precursor solution followed by toluene drop-casting, which led to uniform and dense FASnI 3 perovskite layers. The role of DMSO was to retard the crystallization of FAI and SnI 2 during spin-coating process (Figure 11a-d).
The realized smooth and dense perovskite layer enables a maximum efficiency up to 4.8% for FASnI 3 -based PSCs and the encapsulated devices kept stable for over 100 d. The same method was also applied in low-dimensional tin halide perovskites (PEA) 2 (FA) 8 Sn 9 I 28 to obtain compact and smooth perovskite surface morphology. [64] Chlorobenzene, [161] as a common antisolvent applied in lead-based PSCs, was adopted by Zhang et al. [60] as antisolvent to achieve a dense FASnI 2 Br film giving a device PCE of 1.72%. As an antisolvent, diethyl ether seems to work better than chlorobenzene in improving the morphology of Sn-based perovskites. [56,57] For example, recently, Liao et al. [56] used diethyl ether as an antisolvent in solvent engineering process to synthesize uniform and pinhole-free FASnI 3 perovskite thin films. The fresh spin-coated films showed a reddish intermediate state after dripping with diethyl ether, which might be crucial for the good film morphology, in contrast to chlorobenzene and toluene-based antisolvents, which led to black films immediately (Figure 11e-h). Champion PCE of 6.22% and high reproducibility were achieved based on FASnI 3 PSCs, showing good stability for 30 d (maintaining 85% of its initial efficiency stored in dark and glove box) with encapsulated cells. Later, Ke et al. [57] also used diethyl ether as antisolvent to prepare uniform FASnI 3 perovskite films with high surface coverage on the neat TiO 2 and TiO 2 -ZnS substrates. They pointed out that the use of diethyl ether as an antisolvent in fabrication procedure was advantageous to inhibiting phase separation caused by excess SnF 2 . They proved again that the film prepared with chlorobenzene as antisolvent exhibited poor film coverage (Figure 11i,j). However, Zhao et al. [28] used chlorobenzene as antisolvent to obtain (FA) 0.75 (MA) 0.25 SnI 3 and FASnI 3 film with complete coverage and no phase separation. The exceptionally good effect of chlorobenzene may be due to the use of different solution solvent. This conjecture needs to be proved by more studies.
Solvent-solvent extraction technique is a derivative method of solvent engineering, which was first proposed by Zhou et al. [168] for the fabrication of high-quality lead perovskite thin films. The same method was applied by Milot et al. [171] in Sn-based perovskites with DMF and DMSO as mixed solvents for perovskite precursor solution. The wet film was immediately immersed into an antisolvent (anisole) after spin-coating, producing a smooth, and continuous FASnI 3 thin film. An appropriate crystallization speed is very important to the morphology of the perovskite films. Very recently, Fujihara et al. [48] employed a mixture of toluene and hexane as the antisolvents and DMSO as the good solvent (Figure 12a). Depending on the ratio of antisolvents and temperature, they can control the crystallization speed of the MASnI 3 perovskite films to achieve high surface coverage perovskite films on a planar PEDOT: PSS electrode (Figure 12b

Vapor Deposition
Generally, vapor deposition process is known to provide higher controllability over perovskite films fabrication in terms of higher homogeneity, smoothness, and surface coverage than solution-processed films. [166,172] Moreover, the vacuum condition adopted in vapor deposition process is especially beneficial to air-sensitive tin perovskite. In 2015, Weiss et al. [173] proposed a two-step process combining vapor-deposited SnI 2 precursor films and solution deposited MAI for the preparation of MASnI 3 perovskite films. The results showed homogeneous preformed SnI 2 film and complete conversion of SnI 2 to MASnI 3 . The final film showed complete surface coverage even with such short contact period (Figure 13a-f). In the same year, smooth MASnBr 3 thin films were synthesized via sequential evaporation by Jung et al. [52] (Figure 13g,h). The obtained planar structure device showed an efficiency of 1.12%. In addition, SnF 2 -doped CsSnBr 3 film with excellent ambient air stability was also prepared by sequential vapor deposition method. [75] Later, a hybrid thermal evaporation method at room temperature for the fabrication of high-quality MASnI 3 perovskite thin film was reported by Yu et al. [50] The as-deposited MASnI 3 thin films have excellent morphology, with smooth surfaces, high surface coverage, and strong crystallographic preferred orientation along the <100> direction. Inverted planar architecture solar cells devices were fabricated based on these films and gave an open-circuit voltage up to 494 mV.

Vapor-Assisted Solution Process
The so-called vapor-assisted solution process (VASP) was first reported to construct a high-quality MAPbI 3 film by Chen et al. [174] in 2013. To improve tin perovskite surface coverage, Yokoyama et al. [135] developed low-temperature vapor-assisted solution process (LT-VASP), a kinetically control gas-solid Adv. Sci. 2018, 5, 1700331 Figure 13. SEM images of a) 100 nm vapor-deposited SnI 2 , nonannealed MASnI 3 perovskite films prepared by spin-coating of b) 6 mg mL −1 , c) 10 mg mL −1 , d) 20 mg mL −1 , e) 40 mg mL −1 MAI solutions, and a f) MASnI 3 perovskite film prepared by use of 20 mg mL −1 MAI and subsequent annealing at 80 °C for 10 min; Reproduced with permission. [173] SEM image of coevaporated g) MASnBr 3 (MABr:SnBr 2 = 4:1) Reproduced with permission. [52] Copyright 2015, RSC and h) MASnI 3 . Reproduced with permission. [50] Copyright 2016, RSC. reaction method, to prepare lead-free MASnI 3 thin films. They pointed out that the substrate temperature (60−80 °C window) of preformed solid SnI 2 is very important for achieving homogeneous and high surface-coverage perovskite films. The acquired high-quality MASnI 3 films were fabricated in solar cells with an efficiency of 1.86% with good reproducibility. It is important to point out that LT-VASP method is a pioneer work that explored alternative suitable fabrication methods for tin perovskite films (Figure 14a-c).

Thermal Annealing
Thermal annealing is a critical step in most of the perovskite film deposition steps. It can influence the film formation of perovskites significantly by driving solid-state coarsening of perovskite grains. An interesting work was done by Wang et al. [67] involving an all-inorganic and thermally stable B-γ-CsSnI 3 films deposited by spin-coating. The films were postannealed at different temperatures over a range between 100 and 300 °C for 2 min to coarsen the grains. The B-γ−CsSnI 3 thin films annealed at 150 °C displayed large grain size, high film smoothness, and moderate Sn vacancy (V Sn ) generation, which are responsible for the best performing PSC devices. The B-γ-CsSnI 3 film after 150 °C annealing was applied in an inverted planar device architecture with nickel oxide (NiO x ) as the photocathode. They achieved a PCE of 3.31% without the use of any additive. This work demonstrated that proper thermal annealing is another efficient method for preparing high-performance Sn-based PSCs (Figure 14d-j).

Summary
In summary, as two less toxic family members of Pb, Sn and Ge are deemed as the redeemer to the toxic Pb element and tremendous efforts have been put in optimizing the materials and devices. So far, Sn-based perovskites with lower bandgaps than lead analogs, have obtained a "stable" efficiency up to 6%. In this case, the issue of Sn 2+ oxidization has been partly overcome by adding divalent tin halide additives and some reductive reagents. Additionally, the poor morphology of the tin halide perovskite layer has been improved by the various fabrication methods. Recently, great progress has been made in Sn-based PSCs with inverted device architecture, [28,56,64,66,67] due to the omission of doped HTMs. Unlike Pb-based PSCs where the high efficiencies are usually achieved with doped HTMs, the use of dopants will accelerate the deterioration of tin perovskites. Hence, exploring high-performance and dopant-free HTMs is very important to get efficient and stable Sn-based PSCs. And it is necessary to further study tin-based perovskite material fundamentally (such as the mechanisms of self-doping) and explore novel device structures with different charge selective contact materials toward more efficient and stable Sn-PSCs. For germanium perovskites, the study of these compounds is very rare in the photovoltaic application so far. In this case, due to the relatively small ionic radius of Ge 2+ , the octahedra [GeX 6 ] 3− network is heavily distorted, which leads to wide band gap (>1.6 eV). In addition, the poor solubility of these compounds in polar solvents causes terrible morphology with low efficiency of only 0.2% from solution process. It is urgent to find new preparation methods of germanium perovskites for more efficient Ge-based PSCs. Moreover, the easy oxidation of Ge 2+ and Sn 2+ will be definitely a challenge.

The Group 15 Metals Bi and Sb-Based Absorbers
Beyond group 14 elements, two of group 15 metals in the periodic table, bismuth (Bi) and antimony (Sb) have been also studied for replacing lead in the solar energy absorbing materials. In this section, a brief introduction of typical A 3 Bi 2 X 9 , ABi 3 X 10 , A 3 Sb 2 X 9 (A = MA+, FA+, Cs+; X = I − , Br − , Cl − ) polymorphs and other derivatives will be given. Compared with Sn-based perovskites, they form more diverse dimensionality in terms of the connection type of BiX 6 3− (SbX 6 3− ) octahedron. [101] Here we divided them into 0D, 1D, 2D, and 3D structures.  Reproduced with permission. [135] Copyright 2016, ACS. SEM micrographs of the B-g-CsSnI 3 thin films annealed at different temperatures for 2 min. d) As-deposited, e) 100 °C, f) 150 °C, g) 200 °C, h) 250 °C, and i) 300 °C. j) Average grain size of the B-g-CsSnI 3 thin films with increasing annealing temperature. Reproduced with permission. [67]

Bismuth-Based Absorbers
Being adjacent to Pb 2+ in the periodic table, Bi 3+ has the similar 6s 2 6p 0 electronic configuration, which endows MAPbX 3 with the strong light absorption and long carrier lifetimes. [176] More importantly, it is much less toxic than Pb, [177,178] and has been used in organic synthesis and medicines. [179][180][181] Hence, Bi-based perovskite or hybrid materials are attractive options to replace lead perovskites.

0D Hybrid Materials: MA 3 Bi 2 I 9 Hybrid Bismuth Iodides:
Among all the reported bismuth-based absorbers, organicinorganic hybrid bismuth halide MA 3 Bi 2 I 9 is the most studied polymorph type. Owing to the tervalence state of Bi 3+ , the solid structure of MA 3 Bi 2 I 9 features two face-sharing 0D perovskite structure, [88,182,183] which is constructed by the MA + surrounded binuclear octahedral (Bi 2 I 9 ) 3− clusters, contrasting to the 3D MAPbI 3 perovskite (Figure 15a).
The single crystal of MA 3 Bi 2 I 9 normally displays color like red wine [89] and has regular hexagonal shape with a diameter ranging from 100 to 200 µm, [87] or even up to 4-5 mm, [89] adopting P6 3 /mmc space group. [88,182,184] The optical bandgap of MA 3 Bi 2 I 9 was reported to be ≈1.94-2.26 eV [87][88][89] with absorption coefficient up to ≈1 × 10 5 cm −1 , the same order of magnitude with MAPbI 3 . [185] The valence band maximum (VBM) of the MA 3 Bi 2 I 9 was measured to be 5.63 eV by photoelectron spectroscopy in the air (PESA) [89] and 5.9-6.0 eV by ultraviolet photoelectron spectroscopy (UPS) [87,89] in vacuum, respectively, which are all well aligned with the conduction band of TiO 2 . The electron mobility of MA 3 Bi 2 I 9 single crystal is 29.7 cm 2 V −1 s −1 as estimated by the space charge limited conduction (SCLC), [89] which is comparable to that of MAPbI 3 (38 cm 2 V −1 s −1 ). [186] The same carrier mobility was estimated to be 1 cm 2 V −1 s −1 by Hall Effect. [87] The phase-pure and compact MA 3 Bi 2 I 9 film showed long PL decay over 0.76 ns, with the bulk lifetime approach to 5.6 ns. And this film exhibited robust air stability than MAPbI 3 after 25 d of continuous air exposure with 61% relative humidity (Figure 15b,c). It was speculated that the high air stability is owing to the formation of Bi 2 O 3 or BiOI from BiI 3 , which could serve as a protective layer to prohibit the ingress of water and oxygen into bulk materials. [92] Therefore, on account of its good optoelectronic properties and excellent stability, MA 3 Bi 2 I 9 is the prevailing candidate materials for replacing lead perovskites in the photovoltaic application.
The first report on MA 3 Bi 2 I 9 as light absorbers used in solar cells was by Park et al. [88] They studied the morphology of MA 3 Bi 2 I 9 and Cl-doped (MA 3 Bi 2 I 9 Cl x ) films on TiO 2 substrate by SEM. The results showed that MA 3 Bi 2 I 9 had interconnected layers compared to the more particle-like structure of MA 3 Bi 2 I 9 Cl x . Meanwhile, they pointed out that the excellent surface morphology of MA 3 Bi 2 I 9 could be propitious to form excitons with lower energy. In traditional mesoscopic solar cells, the best device based on MA 3 Bi 2 I 9 showed J SC = 0.52 mA cm −2 , V OC = 0.68 V, FF = 0.33, and η = 0.12%. By contrast, MA 3 Bi 2 I 9 Cl x showed J SC = 0.18 mA cm −2 , V OC = 0.04 V, FF = 0.38, and η = 0.003%. The extreme low V OC of MA 3 Bi 2 I 9 Cl x was attributed to the poor morphology with perovskite particles surrounded by amorphous BiCl 3 . Two months later, Lyu et al. [87] employed poly(3-hexylthiophene-2,5-diyl) (P3HT) as the HTM to replace spiro-OMeTAD in the MA 3 Bi 2 I 9 -based solar cells with a mesoscopic structure, which showed a PCE of 0.19% (Figure 16b).
It became a comment sense that the performance of MA 3 Bi 2 I 9 -based PSCs was improving due to the better morphology of the absorber layer. Singh et al. [86] deposited uniform MA 3 Bi 2 I 9 layers atop mesoporous anatase TiO 2 and exhibited best PCE of 0.2%, as well as 10 weeks stability of the device in ambient condition. Zhang et al. [187] used smooth indium tin oxide (ITO)/glass substrate to achieve a dense MA 3 Bi 2 I 9 thin film, which gave a maximum PCE of 0.42% and high FF up to 0.64. This is one of the highest performances among MA 3 Bi 2 I 9based lead-free PSCs.
More efforts have been paid to modulate the morphology of MA 3 Bi 2 I 9 perovskite. For example, 1-methyl-2-pyrrolidinone (NMP) [85] as morphology controller was added into the MA 3 Bi 2 I 9 -DMF precursor solution, producing a homogeneous film of MA 3 Bi 2 I 9 . The device yielded highly reproducible PCE of 0.31% and kept stable for 30 d in an ambient atmosphere (relative humidity of 50-60%). On the other hand, antisolvent assisted crystallization (ASAC) method also was used to improve MA 3 Bi 2 I 9 thin films like the cases in the lead-and tin-based PSCs. Abulikemu et al. [89] first used this method, but gave a PCE of only 0.11%. Very recently, Mali et al. [84] Adv. Sci. 2018, 5, 1700331   Figure 15. a) Crystal structure of (CH 3 NH 3 ) 3 Bi 2 I 9 (MBI) (left) local structure of the Bi 2 I 9 3− anion and (right) cation and anion positions in the unit cell. Reproduced with permission. [183] Copyright 2016, RSC. Air stability of MBI, measured from mid-July to mid-August in Cambridge, MA, USA. b) Photographs of MBI and MAPbI 3 on quartz over time in the ambient air. c,d) Normalized XRD patterns of MBI over time with air exposure. e) The relative change in the normalized intensity of the diffraction peaks of MBI (day 25 vs. day 1). Reproduced with permission. [92] achieved cuboid-shaped crystals of MA 3 Bi 2 I 9 on the surface of mesoporous TiO 2 by this method. The obtained thin film had excellent air stability with almost no color change even after two month exposure to air. The best solar cell devices made from this kind of films exhibited a PCE of 0.36%, with no substantial efficiency loss after 60 d. Gas-assisted deposition method was first reported by Huang et al. [188] to create uniform and dense lead perovskite thin films. Naturally, this method was used [90] to improve the quality of MA 3 Bi 2 I 9 films. Ultimately a PCE of merely 0.08% was obtained, which is 17% higher compared with the conventional one-step method. Besides the common n-i-p-type devices, the first p-i-n planar heterojunction device of MA 3 Bi 2 I 9 perovskite was reported by Öz et al. [91] with PCE of ≈0.1%. To achieve a smooth, uniform, and compact MA 3 Bi 2 I 9 film in the p-i-n device structure, Ran et al. [83] reported a twostep (evaporation and spin-coating) process of MA 3 Bi 2 I 9 and obtained a PCE of 0.39% and V OC as high as 0.83 V, which is the highest V OC among MA 3 Bi 2 I 9 -based solar cells so far (Figure 16d).
Sulfur-Doped MA 3 Bi 2 I 9 : Sulfur-doped MA 3 Bi 2 I 9 was developed [93] to reduce bandgap of MA 3 Bi 2 I 9 (2.1 eV), which is a relatively higher for the ideal single junction solar cell. [129] Sulfur-doped bismuth perovskites were obtained by in situ sulfur doping of MA 3 Bi 2 I 9 through the thermal decomposition of Bi(xt) 3 (xt = ethyl xanthate) precursor. The color of obtained perovskite films changed from orange to black when annealed from 80 to 150 °C, and there was a notably red shift in the optical absorption edge (Figure 17a). The bandgap of sulfur-doped bismuth perovskite was measured to be 1.45 eV, which is even lower than the prototype MAPbI 3 . [189] Moreover, sulfur-doped MA 3 Bi 2 I 9 exhibited a high carrier mobility of 2.28 cm 2 V −1 s −1 , about twice than that of pristine MA 3 Bi 2 I 9 . This work showed that doping could reduce the bandgap of MA 3 Bi 2 I 9 while improving the charge transport, and further might enhance the performance of MA 3 Bi 2 I 9 -based solar cells. However, unfortunately, there was no device prepared based on such kinds of material.
(BiI 3 ) 0.8 (MA 3 Bi 2 I 9 ) 0.2 : One of the problems with MA 3 Bi 2 I 9 is that its bandgap is even wider than that of BiI 3 (1.8 eV). Therefore, to improve light absorption of MA 3 Bi 2 I 9 , Lan et al. [94] designed active composite layers taking advantages of optoelectronic properties of BiI 3 [190][191][192][193] and suitable energy level alignment of MA 3 Bi 2 I 9 with TiO 2 (Figure 17b). When 20% of MA 3 Bi 2 I 9 perovskite was introduced into the active layers, the (BiI 3 ) 0.8 (MA 3 Bi 2 I 9 ) 0.2 solar cells displayed improved V OC from 0.44 to 0.57 V, with a PCE of 0.08%.
A 3 Bi 2 I 9 (A: Cesium, Formamidinium, Imidazolium, Cyclohexyl Ammonium): Replacing MA with formamidinium (FA), the FA 3 Bi 2 I 9 [95] exhibits the same structure as MA 3 Bi 2 I 9 , with a bandgap of 2.0 eV. Additionally, when more bulky cations like imidazolium and cyclohexyl ammonium are used as the A cation, 0D perovskite-like structure are formed. The synthesized (C 3 H 5 N 2 ) 3 [Bi 2 I 9 ] [96] has two temperature induced solidsolid structural phase transition, and (C 6 H 14 N) 3 Bi 2 I 9 [97] has red emissions at room temperature. All-inorganic bismuth halide compounds have also been studied to replace lead perovskites in PSCs. Cs 3 Bi 2 I 9 as a light harvester was first studied by Park et al. [88] with face-sharing octahedra dimer ((Bi 2 I 9 ) 3− , P6 3 /mmc space group, 0D structure) similar to MA 3 Bi 2 I 9 . It possesses a bandgap of 2.2 eV close to MA 3 Bi 2 I 9 (2.1 eV). A detailed study of its band gap structure was shown by Zhang et al. [194] Compared with MA 3 Bi 2 I 9 and MA 3 Bi 2 I 9 Cl x , Cs 3 Bi 2 I 9 displays a Adv. Sci. 2018, 5, 1700331   Figure 16. a) Large 70 mm thick single crystal of (CH 3 NH 3 ) 3 Bi 2 I 9 grown on the ITO substrate. Reproduced with permission. [89] Copyright 2016, RSC. J-V curve of b) the P3HT based n-i-p device. Reproduced with permission. [87] Copyright 2016, Springer. And c,d) p-i-n devices. c) Reproduced with permission. [91] Copyright 2016, Elsevier. d) Reproduced with permission. [83] Copyright 2017, ACS.

2D Hybrid Materials: A
A 3 Bi 2 Br 9 (A: MA, Cs): Besides the change of A cations, when the halide atom is changed to Br − , one can change the typical 0D Cs 3 Bi 2 I 9 to 2D layered perovskites. For example, MA 3 Bi 2 Br 9 [102] crystallizes into trigonal symmetry ( 3 P m1) space group, forming corrugated layers of BX 6 octahedra (Figure 19b). MA 3 Bi 2 Br 9 possesses a direct bandgap of 2.5 eV [102] and emits at 430 nm as quantum dots with photoluminescence quantum yield (PLQY) up to 12%. Another layered inorganic halide bismuth compound, Cs 3 Bi 2 Br 9 was reported by Bass et al. [199] It occupies the 2D structure, with corrugated layers of cornersharing BiBr 6 3− octahedra, as illustrated in Figure 19d, different from iodine analogs with 0D structure. Cs 3 Bi 2 Br 9 has a large exciton binding energy of 940 meV, which is indicative of a strongly localized character and resulted in highly structured Adv. Sci. 2018, 5, 1700331   Figure 17. a) XRD patterns and color for sulfur-doped MBI at different postheating temperatures. Reproduced with permission. [93] Copyright 2016, ACS. b) Color of (BiI 3 ) 1−x (MBI) x films prepared by the solution method. Reproduced with permission. [94] Copyright 2017, Elsevier. c) J-V curves MA 3 BiI 2 I 9 Cl x , MA 3 Bi 2 I 9 , and Cs 3 Bi 2 I 9 -based solar cell devices. Reproduced with permission. [88] emission. However, this large exciton binding energy will extremely limit its application in photovoltaic technologies, particularly as a light harvester.
CsBi 3 I 10 : Johansson et al. [100] reported another type of cesium bismuth iodine compound CsBi 3 I 10 . In contrast to previously reported Cs 3 Bi 2 I 9 , CsBi 3 I 10 has a different orientation of crystal growth, which may explain a more uniform and smoother coverage on TiO 2 . CsBi 3 I 10 film possesses a smaller bandgap of 1.77 eV and higher absorption coefficients up to 1.4 × 10 5 cm −1 , which are advantageous to PV application, compared with the bandgap of Cs 3 Bi 2 I 9 at 2.03 eV and absorption coefficients of 7 × 10 4 cm −1 (Figure 19e). For the same reason, this material was also used in a red-light photodetector recently. [200] The PV device with a structure of glass/FTO/compact TiO 2 /mesoporous TiO 2 /CsBi 3 I 10 /P3HT/Ag, showed a PCE of 0.4%, with a notable J SC of 3.4 mA cm −2 . This work proved the possibility to further Figure 18. Top-view SEM image of a) HDABiI 5 and b) MAPbI 3 deposited on mTiO 2 /cTiO 2 /FTO. Reproduced with permission. [198] Copyright 2016, RSC.
(TMP) 1.5 [Bi 2 I 7 Cl 2 ]: It is interesting that (TMP)[Bi 2 I 5 ] with one kind of halide species show 1D chain structure, while the addition of Cl leads to the formation of (TMP) 1.5 [Bi 2 I 7 Cl 2 ] with a 2D structure. [101] It has an optical bandgap of 2.10 eV and improved electrical conductivity of 2.37 × 10 −6 S cm −1 . Moreover, it displayed efficient photoconductivity response and very high stability either in humid air or long-time irradiation in a simple device.

3D Hybrid Materials (Double Perovskites):
The organicinorganic hybrid bismuth perovskites mentioned above are all low dimensional structures with PSCs efficiency of only ≈1%. To realize the 3D perovskite architecture which has demonstrated advantages for high efficiency in lead perovskites, double perovskite with 3D structure was developed. Incorporating a monovalent metal into bismuth perovskites could yield a 3D double perovskite with the chemical formula of A I 2 B I Bi III X 6 . Double perovskites usually exhibits a high tolerance to defects owing to the strong ionic nature of the constituents and 3D structure similar to organolead halide perovskites. [201] But they usually have large bandgaps that prevent absorption of the whole solar spectrum. [38,109,202,203] MA 2 B I Bi III X 6 (X: I, Br, Cl; B: Tl, K, Ag): Early in 2015, Giorgi et al. [175] proposed double-perovskite structure MA 2 TlBiI 6 computationally, by substituting Pb 2+ with Tl + and Bi 3+ from parental MAPbI 3 , but stayed in theory because monovalent metal Tl is very toxic. As a compromise, the toxic thallium can be replaced by other monovalent metals like potassium. Therefore, organic-inorganic hybrid double-perovskite MA 2 KBiCl 6 was first synthesized by Wei et al., [104] which is solution processable but with a bandgap of 3.04 eV similar to that of the prototypical MAPbCl 3 perovskite. Because its bandgap is too large to be used for the photovoltaic application. They [105] finally turned to MA 2 TlBiBr 6 , which is isoelectronic with MAPbBr 3 . MA 2 TlBiBr 6 adopts a space group of 3 Fm m like Cs 2 AgBiX 6 (X = Cl, Br) (vide infra) and possesses a narrower bandgap of 2.16 eV than MA 2 KBiCl 6 (Figure 20k). Very recently, [106] they synthesized another new hybrid perovskite: MA 2 AgBiBr 6 with a low bandgap of 2.02 eV and without toxic element. MA 2 A-gBiBr 6 also forms in cubic space group 3 Fm m, with better thermal stability (decomposition temperature up to 550 K) than MAPbBr 3 , and no obvious phase transition was detected from 120 to 360 K. Additionally, the crystal color changes from red to yellowish brown upon cooling (Figure 20j).
Cs 2 A I B III X 6 (A: Ag, (Ag 1−a Tl x ); B: Bi, In, Bi 1−x In x , Bi 1−x Sb x ; X: Cl, Br): Like organic-inorganic hybrid bismuth compounds, inorganic ternary metal halides (A 3 Bi 2 X 9 ) usually occupy low dimensional structures. However, before the report of organic-inorganic hybrid bismuth double-perovskites, all-inorganic halide bismuth double-perovskite (3D) Cs 2 AgBiBr 6 was already synthesized by Slavney et al. [38] Cs 2 AgBiBr 6 has an indirect bandgap of 1.95 eV, and the material shows a long PL decay of 660 ns at room temperature (Figure 20d-f). Additionally, Cs 2 AgBiBr 6 shows higher stability with heat and moisture than MAPbI 3 . Almost the same time, McClure et al. [109] reported Cs 2 AgBiBr 6 with the chloride analogue Cs 2 AgBiCl 6 , with indirect bandgap of 2.19 and 2.77 eV, respectively. Both compounds adopt the cubic doubleperovskite structure with space group 3 Fm m. They are stable when exposed to air, but with the additional light, Cs 2 AgBiBr 6 degrades over a period of weeks. In contrast, Cs 2 AgBiCl 6 shows better stability with no apparent change observed. Volonakis et al. [202] also designed and synthesized Cs 2 AgBiCl 6 (Figure 20g-i). The bandgaps of double perovskites were between 1.95 and 3.04 eV, which were too wide to be used as absorbers in single junction photovoltaic cells. To lower bandgap, Slavney et al. [111] incorporated Tl + as a dilute impurity into Cs 2 AgBiBr 6 , achieving an opaque black octahedral perovskite crystals Cs 2 (Ag 1−a Bi 1−b )Tl x Br 6 (0.003 < x = a + b < 0.075) with very stable structure. The Tl-doped compounds Cs 2 (Ag 1−a Bi 1−b )Tl x Br 6 displayed low bandgap down to 1.40 eV (indirect) and 1.57 eV (direct) when x = 0.075, which is competitive with that of MAPbI 3 . [16] Moreover, time-resolved photo conductivity measurements showed that the Tl-doped materials had long-lived carriers up to microsecond, though shorter than that of Cs 2 AgBiBr 6 due to the extra doping of Tl. This study demonstrated the first double perovskite that has comparable band gap and carrier lifetime to those of MAPbI 3 , but regrettably, there was still toxic Tl in the compounds (Figure 20l). Additionally, through alloying of trivalence In III /Sb III into Cs 2 AgBiBr 6 , the bandgap of double perovskite Cs 2 Ag(Bi 1−x M x )Br 6 [112] (M = In or Sb) can be modulated. For example, when M is Sb III and x = 37.5%, Cs 2 Ag(Bi 0.625 Sb 0.375 )Br 6 had a bandgap of 1.86 eV, which is 0.41 eV lower than the previous ternary compound.
To obtain an absorber with a direct bandgap, Volonakis et al. [113] replaced Bi with In and calculated the band structure of Cs 2 AgInX 6 (X = Cl, Br, I) by first-principles calculations. The combined experiments identified that Cs 2 InAgCl 6 has a direct bandgap of 3.3 eV. The potential of A 2 B′B″X 6 type double perovskites for PV application was further studied by theoretical methods, [105,202,204] and focus was put on A 2 In + Bi 3+ X 6 perovskites. For example, Zhao et al. proposed Cs 2 InBiCl 6 and Cs 2 InSbCl 6 with low direct bandgaps ≈1 eV by HSE+SOC calculation. [204] However, Xiao et al. [205] used a combination of theoretical and experimental study to show that Cs 2 InBiCl 6 and Cs 2 InSbCl 6 were unstable due to spontaneous oxidation of In + into In 3+ . Lately, Volonakis et al. [206] proposed the rule of designing a useful double perovskite material: mimicking the electronic structure of MAPbI 3 . To stabilize the double perovskite structure of A 2 In + Bi 3+ X 6 , they suggested the use of mixed-A-site-cation double perovskite (Cs/MA/FA) 2 InBiBr 6 rather than all-inorganic double perovskites. Although their attempts to synthesize MA 2 InBiBr 6 and FA 2 InBiBr 6 were failed, more efforts are needed to explore the suitable composition of the cations.
Among the reported double-perovskites, Cs 2 AgBiBr 6 is the only one that was applied in a working device. Very recently, Greul et al. [110] prepared phase pure Cs 2 AgBiBr 6 films with a optimal post-annealing temperature of 285 °C. The corresponding mesoscopic devices displayed an incredible maximum PCE of 2.43%, with JSC = 3.93 mA cm −2 , V OC = 0.98 V and FF = 0.63. Moreover, stability of Cs 2 AgBiBr 6 -based devices was tested under constant illumination at ambient conditions during 100 min. This work suggested the potential of doubleperovskites as lead-free alternatives to MAPbI 3 . Figure 20. a) Single-crystal structure of AgBi 2 I 7 cubic structure six-coordinated silver-iodide octahedron sites. b) Tauc plot of AgBi 2 I 7 from the UV/Vis spectroscopy to determine E g under the assumption of a direct bandgap. c) J-V curves in the dark and illumination under 100 mW cm −2 AM 1.5 G. Reproduced with permission. [114] d) Single-crystal structure of Cs 2 AgBiBr 6 ; Photograph of the single crystal; The Bi 3+ face-centered-cubic sublattice, consisting of edge-sharing tetrahedra. e) Absorbance spectrum of Cs 2 AgBiBr 6 powder; f) Time-resolved room-temperature PL and fits for the PL decay time (t) in powder and single-crystal samples. Reproduced with permission. [38] Copyright 2016, ACS. g) Refined crystal structure of Cs 2 AgBiCl 6 ; Diffuse reflectance spectra for h) Cs 2 AgBiBr 6 and CH 3 NH 3 PbBr 3 and i) Cs 2 AgBiCl 6 and CH 3 NH 3 PbCl 3 ; Reproduced with permission. [109] Copyright 2016, Springer.Crystal structure of (j) (MA) 2 AgBiBr 6 ; Reproduced with permission. [106] Copyright 2017, Springer. And k) (MA) 2 TlBiBr 6 ; l) Photographs of Cs 2 AgBiBr 6 and Cs 2 (Ag 1−a Bi 1−b )Tl x Br 6 (x = a + b = 0.075) single crystals and change of absorption onset. Reproduced with permission. [111] Copyright 2017, ACS.

www.advancedscience.com
Adv. Sci. 2018, 5,1700331 AgBi 2 I 7 and Ag 2 BiI 5 : Besides double-perovskites, Kim et al. [114] synthesized AgBi 2 I 7 with cubic-phases composed of vacancyfree corner-sharing bismuth iodide hexahedra and silver iodide octahedra. (Figure 20a-c) They fabricated dense, smooth, and pinhole-free AgBi 2 I 7 thin films with 200-800 nm large grains after annealing at 150 °C. The AgBi 2 I 7 film absorbs light across the range from 350 to 750 nm, with an E g value of 1.87 eV. They applied it in solar cells and the best AgBi 2 I 7 -based device had a PCE of 1.22 % and showed good stability with only 8% PCE reduction over 10 d under ambient conditions. From solutions with different ratios of AgI and BiI 3 (AgI/BiI 3 = 2:1), Zhu et al. got a new crystal structure of Ag 2 BiI 5 with a space group of R3 m. [115] The Ag 2 BiI 5based devices showed a maximum IPCE of 45% and a promising PCE above 2%. The results show the potential of finding new leadfree absorbers and the possibility to tune the properties of bismuth halides by adding a different ratio of precursors.

Antimony-Based absorbers
Antimony (Sb) is on the top right-hand corner of lead in the periodic table, and its trivalent cation possesses a similar electronic configuration with divalent Pb 2+ . Antimonial compounds have been studied and used as therapeutic agents for human leishmaniasis and demonstrated low toxicity with appropriate regulations. [207,208] Hence, Sb is expected to be a nontoxic alternative to lead as well. Due to the high oxidation state (+3), Sb 3+ -based halides have crystal structures of low dimensionality with the typical chemical structure A 3 Sb 2 X 9 , forming in dimer structure or layered structures. [119,209] Cs 3 Sb 2 I 9 : Depending on the synthesis conditions, Cs 3 Sb 2 I 9 forms completely different solid structure. From solution preparation, Cs 3 Sb 2 I 9 preferentially forms 0D structure with isolated dimers of face sharing octahedrons (space group P6 3 /mmc, no. 194) similar to that of Cs 3 Bi 2 I 9 . While from solid-state or gas-phase reactions, it forms 2D layered structure ( 3 P m1, no. 164) (Figure 21a,b). In 2015, Saparov et al. [119] reported preparation and characterization of Cs 3 Sb 2 I 9 thin films, and the first solar cell using Cs 3 Sb 2 I 9 as light absorbers. The prepared Cs 3 Sb 2 I 9 derivative from two-step deposition approach has a layered structure and shows large grains above 1 µm. The layered Cs 3 Sb 2 I 9 film shows red color as opposed to the orange color of the 0D Cs 3 Sb 2 I 9 . The film has a bandgap of 2.05 eV, high absorption coefficients up to 10 5 cm −1 , an ionization energy of 5.6 eV, and better stability in ambient air than MAPbI 3 films Figure 21. a) Removal of every third Sb layer along the 〈111〉 direction of a) the perovskite structure results in b) the 2D layered modification of Cs 3 Sb 2 I 9 ; c) Bandgap of the layered modification of Cs 3 Sb 2 I 9 (inset shows a thin film) using the Tauc relation. Reproduced with permission. [119] Copyright 2015, Springer. d) Schematic showing the influence of A cation size on the structure of A 3 Sb 2 I 9 ; e) J-V curve under forward and reverse scans of the best device with the energy levels of Rb 3 Sb 2 I 9 shown in inset; Reproduced with permission. [35] Copyright 2016, Springer. f) Crystal structure of (CH 3 NH 3 ) 3 Sb 2 I 9 ; g) Comparison of the absorption coefficient of various Bi-based perovskites and (CH 3 NH 3 ) 3 Sb 2 I 9 determined by PDS measurements; h) J-V curve of (CH 3 NH 3 ) 3 Sb 2 I 9 solar cell measured with "up" and "down" sweep with a rate of 0.1 V s −1 . Reproduced with permission. [117] Copyright 2016, ACS.
( Figure 21c). Unfortunately, the PV device with an architecture of glass/FTO/c-TiO 2 /Cs 3 Sb 2 I 9 /PTAA/Au showed PCE below 1% and a low open-circuit voltage between 0.25 and 0.3 V. They ascribed the low performance of Cs 3 Sb 2 I 9 to the deep defects that promote nonradiative recombination, which is substantially suppressed in MAPbI 3 with shallow defects. Later, 0 D structure was also synthetized to fabricate solar cell devices and displays champion efficiency of 0.84% with improved Voc of 0.60 V. [118] Rb 3 Sb 2 I 9 : As mentioned above, Cs 3 Sb 2 I 9 is inclined to form dimer type 0D structure when obtained via a solution process. However, in 2016, Harikesh et al. [35] found that when Cs is replaced by Rb, the resulting compound can easily form layered structures via solution processing (Figure 21d). They compared the formation energies of the dimer and layered forms of A 3 Sb 2 I 9 with A = Cs and Rb by DFT calculations. The results showed that Rb 3 Sb 2 I 9 had higher preference for the layered phase, which is linked to the smaller ionic radius of Rb (1.72 Å), as compared to that of Cs (1.88 Å). Moreover, Rb 3 Sb 2 I 9 is thermally stable up to 250 °C, and no phase transition between −40 to 200 °C, which is beneficial for operating in a solar cell. Additionally, they obtained a near ideal stoichiometry Rb 3 Sb 2 I 9 film with a single 3+ oxidation state of Sb by an excess SbI 3 treatment method. The perovskite films with the SbI 3 treatment showed better coverage compared to the pristine films, with an absorption coefficient of >1 × 10 5 cm −1 and an indirect bandgap of 2.1 eV. The solar cells based on Rb 3 Sb 2 I 9 with poly-TPD as HTM exhibited J SC = 2.11 mA cm −2 , V OC = 0.55 V and a PCE of 0.66% (Figure 21e). MA 3 Sb 2 I 9 : Unlike inorganic Rb 3 Sb 2 I 9 and Cs 3 Sb 2 I 9 , which tend to form two different structures (layered or dimer) depending on the crystallization conditions, when organic cation MA + is used, the hybrid antimony-based perovskites MA 3 Sb 2 I 9 only forms 0D dimer structure, with octahedral anionic metal halide units (Sb 2 I 9 ) 3− surrounded by (MA) + cations [210] (Figure 20f). MA 3 Sb 2 I 9 used in the photovoltaic application was first reported by Hebig et al. [117] They prepared flat and homogeneous thin films of MA 3 Sb 2 I 9 by a two-step spin-coating process followed by toluene treatment. The MA 3 Sb 2 I 9 thin film shows a peak absorption coefficient (α) above 10 5 cm −1 and an optical bandgap of 2.14 eV. Additionally, they found that the Sb-perovskite showed no exciton peak in its absorption spectrum contrasting with the related Bi compound. The Urbach tail energy of this amorphous compound is close to 62 meV, indicating a high degree of energetic disorder, which will bring additional sources of nonradiative recombination engendering low open-circuit voltages. [211][212][213] They fabricated a planar heterojunction solar cell with the architecture of ITO/ PEDOT: PSS/absorber/PC 61 BM/ZnO-NP/Al, which showed a PCE of η ≈ 0.5%, V OC of 0.89 V, and extremely low photocurrent densities (Figure 20h). Very recently, Boopathi et al. [118] reported one-step prepared MA 3 Sb 2 I 9 through optimization of the precursor solution and HI concentrations. The highly crystalline MA 3 Sb 2 I 9 film along with improved surface morphology contributed to the record PCE of 2.04% based on a planar architecture devices.
(NH 4 ) 3 Sb 2 I x Br 9−x : An even larger A cation was used by Zuo and Ding to synthesize a family of layered perovskite type light absorbers (NH 4 ) 3 Sb 2 I x Br 9−x (0 <x < 9). [116] These materials show good solubility in ethanol, variable absorption onset from 558 to 453 nm and high carrier mobility (hole of 4.8 cm 2 V −1 s −1 and electron of 12.3 cm 2 V −1 s −1 ). (NH 4 ) 3 Sb 2 I 9 solar cells gave an extraordinarily high V OC of 1.03 V and a PCE of 0.51%. Cs 4 CuSb 2 Cl 12 : Aiming for high-performance lead-free metal-halide perovskite, Vargas et al. [121] incorporated Cu 2+ into α-Cs 3 Sb 2 Cl 9 and yielded layered perovskite Cs 4 CuSb 2 Cl 12 , which has a direct bandgap of 1.0 eV and better conductivity than that of MAPbI 3 . Additionally, Cs 4 CuSb 2 Cl 12 displayed high thermal-stability, photostability, and resistance to humidity. These properties show that Cs 4 CuSb 2 Cl 12 is a promising material for photovoltaic applications.

Summary
Aforementioned Bi 3+ and Sb 3+ -based perovskites and related absorbers represent the efforts of the scientist to find alternative lead-free active materials in PSCs. Bi 3+ -based compounds are much less toxic, even than divalent Sn 2+ and Ge 2+ ions, while displaying admirable air stability among all metal-halide hybrid absorbers. However, the trivalence oxidation state of Bi makes the related compounds usually form low dimensional phases, with large or indirect band gap (≈2 eV), high exciton binding energy (70-300 meV), [88] and relatively low charge transport ability. On account of the disadvantages, the highest PCE of Bi-based PSCs reported so far is only 2.1%. Though the heterovalent substitution with monovalent metal could yield a 3D double perovskites, they usually display wide bandgap or need to involve toxic elements. Actually, there is no real application of these 3D double perovskites in solar cells to date. A recent theoretical study indicates that these structural 3D double perovskites do not have a 3D electronic structure. [203] Another problem is the poor morphology of Bi-based perovskite, which might originate from the preferred tendency to form regular hexagonal crystalline phase. Thus, two main strategies may be used by the community to construct ideal high-efficiency Bi-based PSCs: (i) compositionally engineered bismuth perovskite with 3D electronic structure and therefore low bandgap and (ii) new film fabrication methods which are suitable for Bi-based perovskite. At the same time, the development of Sb 3+ -based perovskites is still in its infancy. Owing to the trivalence oxidation state similar with Bi 3+ , Sb 3+ -based perovskites also have wide bandgap and low dimensional structure with low efficiency. Sb 3+ -based perovskite has deep level defects (versus shallow levels in MAPbI 3 [214] ), which is extremely detrimental to solar cells performance. Besides, Sb 3+ -based perovskites (Cs 3 Sb 2 I 9 , Rb 3 Sb 2 I 9 ) are prone to form in 0D dimer structure with poor charge transport when they are prepared via a solution process, which is also disadvantageous to high-efficiency solar cells. Though these research results on the photovoltaic performance are unsatisfactory, they paved the way to lead-free halide hybrid absorbers for photovoltaic applications. The dimensionality variation of bismuth and antimony-based absorbers is shown in Table 2.
Adv. Sci. 2018, 5, 1700331 daily life. Unlike conventional Pb-based perovskites with 3D structure, Cu-based perovskites usually form 2D layered structure, due to its smaller ionic radii. Their general formula is (RNH 3 ) 2 CuX 4 , where R-NH 3 + is aliphatic or aromatic ammonium cation and X is a halogen. [122] Cu 2+ with an electronic configuration of 3d 9 (t 2g 6 e g 3 ), is more stable in the air than other two divalent Sn 2+ and Ge 2+ (Figure 22a). In 2015, Cui et al. [122] synthesized two cupric bromide hybrid perovskites, (p-F-C 6 H 5 C 2 H 4 -NH 3 ) 2 CuBr 4 , and (CH 3 (CH 2 ) 3 NH 3 ) 2 CuBr 4 , with absorption from 300 to 750 nm and studied their photovoltaic performance. This is the first report on Cu-based PSCs and showed PCEs of 0.51% and PCE of 0.63%, respectively. Both devices exhibited good air stability with less than 5% decrease of the efficiencies after 1 d in the air with humidity of 50% without encapsulation. Later, Cortecchia et al. [39] reported Cl-doped MA 2 CuBr 4 perovskites, and found that the Cl was essential for stabilizing MA 2 CuCl x Br 4−x perovskites against copper reduction and enhancing the perovskite crystallization. By tuning the Br/Cl ratio, the optical absorption can be adjusted and extended to the near-infrared. Further optimizing the infiltration of mesoporous TiO 2 by 2D copper perovskites yielded a PCE of 0.017% using MA 2 CuCl 2 Br 2 as the light harvester ( Figure 22b). Moreover, Li et al. [123] claimed the syntheses of a highly stable C 6 H 4 NH 2 CuBr 2 I compound by equimolar reaction of hydrophobic C 6 H 4 NH 2 I (2-indoaniline) with low-toxic CuBr 2 . The XRD patterns of the C 6 H 4 NH 2 CuBr 2 I thin film showed almost no change after 4 h of immersion in water, and the printable mesoscopic solar cell based on carbon back-contact achieved the best PCE of 0.46% (Figure 22c). The low device performance is due to low absorption coefficient (<10 5 cm −1 ), [39] anisotropic charge transport in low-dimensional structure and heavy mass of the holes. Moreover, the existence of Cu 2+ reduction could introduce higher trap density, which is unfavorable to photovoltaic performance.

Conclusion and Outlook
We have thoroughly reviewed a series of lead-free halide hybrid absorbers with various metallic cations including Sn, Ge, Bi, Sb, and Cu etc. in the context of solar cells application. It has been proved that the variation of halide elements in the X position of a typical perovskite material will change the E g significantly and the trend follows the electronic negativity of the  [39] Copyright 2016, ACS.

www.advancedscience.com
Adv. Sci. 2018, 5, 1700331 halide ions. [3] Similarly, the variation of A cations also influences the E g in various ways depending on the type of central metal ion [3] (Figure 23a). In the case of lead halide perovskite, the E g is decreasing with the increase of radii of A cations. However, in the case of Ge halide perovskite, the trend is reversed with CsGeI 3 has the smallest E g . For Sb and Bi-based absorbers, the lowest E g appears when Cs + is used as the A cation. The E g will increase no matter the radii of A cation are larger or smaller than Cs + . This phenomenon may be due to the change in the dimensionality of the material. In the case of Sn-based material, MASnI 3 has the lowest E g . Among all the metal cations under study, Sn perovskite has the lowest E g [132] while Ge perovskite gives the highest E g . [95] The differences in E g reflect directly in the short current density (J SC ) of the corresponding devices. As can be seen from Figure 22b, Sn-based perovskites provide the highest J SC , while materials based on Bi and Sb yield the lowest J SC due to the low dimensionality and wide E g . A comparison of the device performance between different lead-free absorbers is visualized in Figure 23c. It is clear that there is still a huge gap in PCEs between them and Pb-based perovskite. Hence, a wise decision should be made while efforts should be put in the most plausible direction.
As the most studied lead-free perovskites, Sn-based absorbers with the retaining 3D framework like Pb analogs hold excellent optoelectronic properties, especially the narrow bandgaps and high carrier mobilities. However, the notorious "self-doping" effect impedes their further development. To suppress "self-doping" effect, various tin halide additives and organic reducing agents were introduced into the active layers. As usual, the bad morphology of perovskite films is detrimental to the device performance. Thus, strategies containing the use of additives, solvent engineering, vacuum process, vapor-assisted solution process (VASP), and thermal annealing were adopted to improve the quality of tin perovskites film. Recently, great progress was made in an inverted device architecture. The combination of SnF 2 additive and antisolvent treatment with chlorobenzene gave a new record efficiency of 8.12% with good reproducibility from (FA) 0.75 (MA) 0.25 SnI 3 -based PSCs. More importantly, 80% of PCE retained after 400 h. Among the various types of tin-based perovskites (different A cations), different cations showed different advantages. First, Sn-based perovskites with FA cation showed higher efficiency and stability than MA and Cs counterparts. It has been speculated that FA cation can endow the perovskite with higher formation energies of Sn vacancies [139] and more resistance against oxidation. [133,138] On the other hand, FA-based hybrid perovskites have better solubility than all-inorganic perovskites. According to the summary in Table 1, no spin coating with solvent-engineering was used in Cs-based perovskites due to their poor solubility, but in Figure 23. a) E g versus the type of metal cation; b) J SC versus E g of the absorbers with different metal cation; c) Comparison of the performance of the device between different lead-free absorbers (data were analyzed based on Table 1). the case of FA counterparts, most good results [28,56,57] originated from solvent-engineering resulted in excellent morphology. Second, mixing organic cations at A position seems an effective method to improve devices performance. FA-MA mixed tin perovskite showed V OC up to 0.61 eV, while FA(MA)-BA(PEA) mixed Sn perovskite yielded a 2D structure with improved air stability. Thirdly, Cs-based perovskites have the best thermal stability up to 200 °C. Due to the high valence Sn 4+ cation, Cs 2 SnI 6−x Br x showed the highest environment stability, the PSCs based on which were processed in the air without using any additives.
Despite that the Sn-based absorbers attained a promising efficiency of ≈8%, it is still far from the best Pb-based perovskites. "Self-doping" effect will still be a challenge to all the researchers working on Sn-based perovskite. A deep insight into the mechanisms of "self-doping" effect is crucial for achieving efficiency up to 15% or higher. A more suitable procedure only employs "intermediate agent" that could be removed in the final stage or even without using any sacrificial additives. Successful compositional engineering could help to give efficient and stable compounds similar to Pb-based perovskites. Additionally, high-performance and dopant-free HTMs could also assist in achieving more efficient and stable Sn-PSCs.
For the absorbers based on metals beyond the group 14, owing to the low dimensionality and wide band gaps, the photovoltaic performance of Bi, Sb and Cu-based devices are still unsatisfactory with efficiency ≈2%. More efforts are needed at two possible directions to open up the avenues toward highperformance Bi-based absorbers. The first one is new double perovskites, which may involve a lot of theoretical calculation and the corresponding experimental study. So far, researchers have used Cs 2 AgBiBr 6 to make device with PCE up to 2.43%. The other one is the ferroelectric perovskites represented by Bi-based compounds. As far back as 1956, photovoltaic (PV) effect has been found in oxide perovskite BaTiO 3 which has no lead elements. [216] Afterward, more lead-free oxide perovskites were studied on PV effect, such as BiFeO 3 , [217,218] BiMnO 3 , [219] [KNbO 3 ] 1−x [Ba Ni 1/2 Nb 1/2 O 3−δ ] x [220] and Bi 2 FeCrO 6 , [221] etc. So far, the highest PCE among all oxide perovskites was obtained as 8.1% by using double-perovskite Bi 2 FeCrO 6 . [221] The ferroelectric oxide perovskites often showed exceptionally high photovoltages, which are normally much larger than their bandgaps.
For the transition metal Cu-based absorbers, like 2D lead perovskites, their wide compositional tunability, and increased environmental stability are their intrinsic advantages. For example, one can introduce optoelectronically active organic cations to increase optical absorption cross-section and improve vertical charge transport. Another approach is to make multidimensional (MD) perovskites by mixing the 2D and 3D materials.
Among all the above-mentioned lead-free absorbers, Snbased perovskites possess the most efficient PCE up to 8.12% while other absorbers showed only below 2.1%, although Bibased absorbers showed the highest air stability. Moreover, theoretical calculation [222,223] indicated that efficiencies above 15% could be obtained from MASnI 3 PSCs. Thus, we argue that the Sn-based absorbers are the most promising surrogate for Pb in PSCs. However, we have bear in mind that Sn element is still harmful to the human body in their practical utilization, [224,225] while Bi, Sb, and Cu are more environmentally friendly. Finally, we should think about our initial question: can we get the clean power output from the new hybrid absorbers without the danger of environmental contamination? [226] The answer may need to be found in the future development of new lead-free hybrid light harvesting materials.
During the revision of our manuscript, we noticed that there are two important Sn-based and one bismuth-based absorbers reported recently. Firstly, ethylenediammonium(en) cations were incorporated into FASnI 3 , [62] MASnI 3 and CsSnI 3 , [55] and then form so-called "hollow" {en}ASnI 3 perovskite. For the sake of incorparation of appropriate amount of en cations, the 3D structure of perovskite is retained while perovskite film morphology is significantly improved. Finally, the best-performing solar cells display a high efficiency of 6.63% of {en}MASnI 3 and 7.14% of {en}ASnI 3 , respectively. These results are presented in table 1. Furthermore, Zhang et al. report a novel two-step vacuum deposition procedure to get homogeneous transformation of BiI 3 to MA 3 Bi 2 I 9 for highly compact, pinhole-free, large-grained films. The solar cells realized a record PCE of 1.64% and also a high EQE approaching 60%. [227]