Organometal halide perovskite thin films and solar cells by vapor deposition

Organometal halide perovskites (OHPs) are currently under the spotlight as promising materials for new generation low-cost, high-efficiency solar cell technology. Within a few years of intensive research, the solar energy-to-electricity power conversion efficiency (PCE) based on OHP materials has rapidly increased to a level that is on par with that of even the best crystalline silicon solar cells. However, there is plenty of room for further improvements. In particular, the development of protocols to make such a technology applicable to industry is of paramount importance. Vapor based methods show particular potential in fabricating uniform semitransparent perovskite films across large areas. In this article, we review the recent progress of OHP thin-film fabrication based on vapor based deposition techniques. We discuss the instrumentation and specific features of each vapor-based method as well as its corresponding device performance. In the outlook, we outline the vapor deposition related topics that warrant further investigation.


Introduction
Organometal halide perovskite (OHP) solar cells have emerged as the most promising candidate for the next generation high efficiency solar cell technology that is compatible with low-cost, low-temperature processing, exible substrates, and large-area fabrication using e.g. ultrasonic spray-coating, 1 printing, 2 rollto-roll, 3 and vapor deposition techniques. 4,5 Laboratory scale cells with the highest power conversion efficiency (PCE) of $20.1% were achieved in a short time span of four years, 6,7 which is only a few percent lower than the best single crystalline silicon solar cells.
The term perovskite refers to a category of materials that can be represented by the building block of ABX 3 and adopt a similar crystal structure to oxide perovskites such as calcium titanate (CaTiO 3 ). A few review papers have been published on oxide and halide based perovskites with emphasis on solar cell application. [8][9][10][11][12][13][14][15][16][17][18][19][20] In the particular case of OHPs, the halide anions (X ¼ I, Br, or Cl) and metal cations (B ¼ Pb, Sn) form the BX 6 octahedral arrangement, Fig. 1a. The BX 6 octahedra extend to a threedimensional network in which cations A can be stabilized within the space formed by the eight adjacent octahedra (Fig. 1). 21 The larger cation A (A being larger than B) can be Cs + , 22 methylammonium (CH 3 NH 3 + , MA + ), ethylammonium (CH 3 CH 2 NH 3 + , EA + ), 15 formamidinium (NH 2 CH]NH 2 + , FA + ), 23 or mixed CH 3 NH 3 and 5-aminovaleric acid (5-AVA) cations [(5-AVA) x (-CH 3 NH 3 ) 1Àx ]. 2 The crystallographic stability and probable structure are estimated by considering the Goldschmidt tolerance factor and the octahedral factor. 9,15,16,24 Nevertheless, the determination of chemical and thermal stability of the resultant perovskite structure requires more detailed analysis. 10 CH 3 -NH 3 PbI 3 , the most commonly employed material in OHP solar cells, was reported to have a high absorption coefficient (direct bandgap of $1.55 eV) and high mobilities for electrons (7.5 cm 2 V À1 s À1 ) and holes (12.5-66 cm 2 V À1 s À1 ), i.e. ambipolar nature, resulting in long carrier diffusion lengths (100 nm to 1 mm). 25 Although the amount and role of incorporated Cl are still under debate, 26 mixed methylammonium-lead halide CH 3 NH 3 PbI 3Àx -Cl x is another type of halide perovskite reported with an even higher charge-carrier mobility ($33 cm 2 V À1 s À1 ), resulting in carrier diffusion lengths of up to 3 mm. 27 Theoretical studies have shown that most point defects in OHP form shallow defect states. [28][29][30][31][32] In addition, grain boundaries were shown that they do not generate gap states, which makes the electronic property behavior of polycrystalline halide perovskite similar to that of a thin-lm single crystal. 28,29,33 In a recent work by deQuilettes et al., 34 the existence of large spatial variations in photoluminescence (PL) intensity and carrier recombination lifetimes were probed using a confocal PL microscope. In particular, higher contrasts were observed at the grain boundaries in comparison to the bulk of the material within the individual grains of CH 3 NH 3 PbI 3Àx Cl x perovskites. 34 Differences in PL intensities were attributed to the variations in radiative and nonradiative recombination dynamics. 34 The question whether the perovskite solar cell system is excitonic, similar to an organic solar cell, which requires a heterojunction interface to separate electron-hole pairs, or instead photoexcitations spontaneously dissociate into free carriers in the bulk, similar to inorganic solar cells, has not been completely settled. [35][36][37][38][39] On the other hand, a number of studies suggest that exciton binding energies (BEs) in perovskites are in the range of $2-50 meV, 35 and ultrafast interfacial charge-transfer dynamics take place; 36 collectively, the majority of these observations imply that perovskite solar cells are predominantly non-excitonic similar to inorganic solar cells showing relatively low exciton BEs, e.g., Si (15.0 meV), GaAs (4.2 meV), and CdTe, (10.5 meV). 37 In the 1990s, Mitzi and co-workers studied OHPs and discovered desirable physico-chemical properties of these materials mainly for electronic applications. [40][41][42] Almost in parallel, Grätzel and co-workers developed a new class of photovoltaic technology, dye-sensitized solar cells (DSSCs or Grätzel cell) and solid-state DSSCs (ssDSSCs) shown in Fig. 2. 14,43,44 Miyasaka and co-workers were the rst to apply perovskites in DSSCs in 2009. 45 The dye was replaced by methylammonium lead triiodide/tribromide (CH 3 NH 3 PbI 3 and CH 3 NH 3 PbBr 3 ) perovskites in DSSC conguration 13,14 obtaining a PCE of $3.8% and $3.1%, respectively, using the iodine/ triiodide redox liquid electrolyte as the hole-transport material. Due to the high instabilities of the perovskite materials in the electrolytes, ruthenium-based dyes were still the preferable choice. In 2012, Park and co-workers fabricated all solid-state perovskite (CH 3 NH 3 PbI 3 ) solar cells and achieved a PCE of $9.7% and much better durability. 46 The key advance was made possible by replacing the liquid electrolyte with a solid hole transporting layer (HTL) material, Fig. 2. 46 Since then, a myriad of reports have been published exploring the different perovskite materials, various device architectures, and fabrication methods. 14,[47][48][49][50] Both the dye and CH 3 NH 3 PbI 3 assume the function of a sensitizer in which light absorption induces subsequent electron injection into the conduction band of the mesoporous TiO 2 scaffold (electron transport layer, ETL) accompanied by hole injection from the oxidized sensitizer to the highest-occupied molecular orbital (HOMO) of the HTL.
Further charge transport of electron and hole through the external circuit completes the photovoltaic operation. 8 The use of TiO 2 and HTL as selective contacts ensures that photoexcited charge carriers (electrons and holes) are transported in opposite directions. In a separate experiment, Lee et al. 47 showed that a mesoporous (mp-) scaffold made of Al 2 O 3 instead of TiO 2 generated a similar PCE even though Al 2 O 3 is an insulating material. This paradigm has been rationalized by suggesting that perovskite itself is a good electron conductor; if so, no mp-TiO 2 scaffold is necessary at all. This led to a much simpler planar-type device architecture, Fig. 2 51 had demonstrated that CH 3 NH 3 PbI 3 could also act as an efficient hole conductor, which could even eliminate the need for employing the additional HTL layer. However, generally both electron and hole selective contacts contribute to the enhancement of the cell ll factor (FF). In particular, the hole selective contact tends to enhance the open-circuit voltage (V oc ) by minimizing interfacial charge recombination processes, i.e. the HTL performs both functions of blocking electrons as well as transporting holes efficiently. 52 The lm morphology, thickness, stoichiometry, crystallinity as well as material purity have signicant impact on the overall solar cell performance. A variety of solution-and vapor-based OHP deposition techniques have been reported including one-step spin-coating, 53-57 two-step deposition techniques, 41,58-60 solventsolvent extraction, 61 vapor-assisted solution processes, 62-69 dualsource vacuum deposition, 4,5,70-77 hybrid deposition, 78-81 hybrid chemical vapor deposition, 82-87 sequential vapor deposition, [88][89][90][91][92] ash evaporation, 93 etc. One-step spin-coating is one of the widely used methods because of its simplicity and low-cost. However, the lms prepared by this method oen have a poor morphology (incomplete coverage) especially in the case of planar architecture, which results in decreased solar cell performance. 54,[94][95][96] In the two-step process, 41,58,59 a layer of metal halide is deposited by spin-coating followed by dipping the lm into the organic salt solution and perovskite formed by a chemical reaction. However, due to the high reaction rates of perovskite formation, it is challenging to optimize the processing conditions with sufficient reproducibility. 97 Despite the fact that laboratory record efficiencies have been obtained by solution processing, 7,50,98 it is observed that the reaction kinetics need to be rigorously controlled to maintain consistent device performance and minimize batch-tobatch variations. 99 Yang and co-workers introduced a CH 3 NH 3 I vapor-based approach for the deposition of a perovskite layer called vapor-assisted solution process (VASP). 62,63,65 In their process, PbI 2 lms were annealed in MAI vapor at 150 C in an N 2 environment for 2 h, Fig. 3a. Perovskite lms exhibited high crystallinity, uniform surface coverage and large grain sizes up to 1 micrometer, Fig. 3b-d. The high quality lms of CH 3 NH 3 PbI 3 enabled enhanced solar cell parameters of short-circuit current (J sc ), V oc , FF, and PCE: 19.8 mA cm À2 , 0.924 V, 0.663, 12.1%, respectively, in a planar architecture, Fig. 3e. 62,63 The surface roughness of the lms was measured by atomic force microscopy (AFM) (5 Â 5 mm 2 , Fig. 3f) and calculated to be 23.2 nm. In a recent publication, the authors state that it is still unclear why the efficiency of perovskite solar cells based on VASP is slightly lower than that of devices derived from an optimized solution process. 65 In this review, we focus on the different vapor-based methods to deposit perovskite lms, which in many cases show properties different from their counterparts prepared by solutionbased methods.

Vapor deposition system description
Vapor deposition techniques are widely used in the semiconductor industry aiming at large scale production in optoelectronic applications. The viability of OHP material synthesis by physical vapor deposition techniques has also been demonstrated. 4,5,[70][71][72][73][74]100 Such techniques offer unique advantages such as (1) it is feasible to fabricate lms with high-  purity as the lms are formed by sublimating the powder precursors aer extensive outgassing under a vacuum environment; (2) in general, the initial nominal stoichiometry of precursors (e.g., CH 3 NH 3 I and PbCl 2 ) can be well controlled in both solution and vacuum evaporation methods. On the other hand, it is necessary to take into account the solubility of precursors in determining the composition of the lms that are prepared by solution methods. For example, it is difficult to dissolve PbCl 2 in N,N-dimethylformamide (DMF) when the CH 3 NH 3 I : PbCl 2 molar ratio is lower than 3 : 1. 47 (3) The commonly used solvents, in the solution process, can get intercalated in perovskite lms. DMF, H 2 O, and dimethylsulfoxide (DMSO) were observed to form stable intermediate complexes of CH 3 NH 3 PbI 3 $DMF, 101 CH 3 NH 3 PbI 3 $H 2 O, 102 and CH 3 NH 3 PbI 3 $DMSO, 103 respectively, likely to affect the perovskite lm stability. (4) Vapor deposition techniques are suitable for the preparation of multilayered structures of thin lms, while it is challenging for solution processing. (5) With proper optimization, perovskite lms can be deposited by vapor deposition on a variety of substrates. The wettability issues in solution processing oen lead to non-uniform coating and pin-hole formation.
In 1997, Era et al. 104 reported for the rst time the dualsource vapor deposition method to form two-dimensional layered hybrid lead iodide intercalated with an organic ammonium layer. The synthesis of (RNH 3 ) 2 PbI 4 layered perovskite was performed under a pressure of $10 À6 Torr sublimating PbI 2 and organic ammonium iodide RNH 3 I (2-phenylethylammonium iodide C 6 H 5 C 2 H 4 NH 3 I was used as RNH 3 I). The synthesis of KPbI 3 under vacuum from the PbI 2 and KI precursor sources was reported by Salau. 105 KPbI 3 has been suggested as a potential candidate for solar cell applications because of its high thermal stability (220 C). 105 In 2013, Liu et al. 4 reported the synthesis of three-dimensional CH 3 NH 3 PbI 3Àx Cl x by using the dual-source vapor deposition technique with PbCl 2 and CH 3 NH 3 I as precursors leading to high efficiency photovoltaic devices (PCE $ 15.4%, Table 1). Similarly, Malinkiewicz et al. deposited a pure CH 3 NH 3 PbI 3 perovskite by using PbI 2 and CH 3 NH 3 I sources showing uniform lm formation with a root mean square (RMS) roughness of 5 nm measured by AFM. 5 In addition, the lms showed uniform grainy structures with an average grain size of 150 nm. 71 The schematic illustration of the dual-source vacuum deposition process is shown in Fig. 4. 71,75 PbX 2 (X ¼ I, Cl) and CH 3 NH 3 I precursor materials contained in crucibles are heated (co-evaporation) to their corresponding sublimation temperatures. CH 3 NH 3 PbI 3Àx Cl x and CH 3 NH 3 PbI 3 perovskites layers are formed on the substrate that is xed at a distance of $20 cm above the crucibles. 5 Typical base pressures of 10 À5 to 10 À6 Torr are reached aer loading the precursor materials. 4,5 The stoichiometry (chemical composition) and lm thickness are monitored with the aid of piezo-electric sensors 4,5 mounted inside the vacuum chamber (or a quartz crystal microbalance, QCM). Because perovskites are formed by the co-evaporation process, it requires the initial calibration as precise as possible for the thicknesses of individual evaporated PbX 2 and CH 3 NH 3 I lms. Material density (r), acoustic impedance (or Z-ratio), and geometric (or tooling factor) are parameters that need to be determined for the calibration of evaporation rate of the material being sublimated. Oen it is difficult to nd those parameters especially for organic compounds. For example, Liu et al. assumed the density and Z-ratio of CH 3 NH 3 I to be 1 g cm À3 and 1, respectively, because its precise density is unknown. 4 The density for CH 3 NH 3 Cl of 1.1 g cm À3 was previously reported. 4 More recently, the density value of 2.224 g cm À3 for CH 3 NH 3 I has been reported. 80 In addition, the tetragonal CH 3 NH 3 PbI 3 perovskite phase was calculated to have a density of 4.149 g cm À3 . 80 The densities of PbCl 2 and PbI 2 can be found in the literature with typical values of 5.85 g cm À3 and 6.16 g cm À3 , respectively. As the source-to-substrate distance generally differs from the source-to-QCM distance, it is oen a common practice to perform some initial tests to determine the tooling factor. A certain amount of material is deposited on a at substrate recording the nominal thickness measured by the QCM with a preset tooling factor value. This nominal thickness value is then compared to the thickness value determined using another technique (e.g., AFM or surface prolometry). The linear relationship provides the new tooling factor of the evaporation system. As it will be discussed in more detail in the next section (3. Hybrid deposition method), the calibration procedure for the CH 3 NH 3 I was reported to be difficult due to the formation of a non-uniform layer dominated by the Volmer-Weber or Stranski-Krastanov growth mode and the volatile nature of the organic lm. 5,71,78,79 Alternatively Malinkiewicz et al. 5,71 kept the evaporation temperature for the CH 3 NH 3 I crucible constant (at 70 C) and varied the CH 3 NH 3 I : PbI 2 ratio by changing only the evaporation temperature of the PbI 2 crucible (250-260 C). The optimum conditions were determined by analyzing the evaporated perovskite lms by grazing incident X-ray diffraction (GIXRD). Once the optimum PbI 2 crucible temperature (250 C) for generating the stoichiometric perovskite is determined, perovskite lms with similar properties can be prepared reproducibly indicating the robustness of the protocol. 5,71 The substrate holder is maintained at near room-temperature during perovskite deposition for the processes described above. 4,5,71 Because of low-temperature processing, it is highlighted that the technique is of particular interest for the deposition of perovskite lms onto exible substrates. Liu et al. 4 provide side-by-side comparison on the morphology of the CH 3 NH 3 PbI 3Àx Cl x perovskite lms prepared by the solution and dual-source vacuum evaporated lms. For example, the top-and side-views of scanning electron microscopy (SEM) highlight that vacuum-deposited lms show full coverage and are extremely uniform with crystalline features on the size scale of hundreds of nanometers, Fig. 5a and b. Large-area cross-sectional SEM images, Fig. 5c and d, reveal that solution-processed lms exhibit large variations in lm thickness (50 to 410 nm) over the sample area, whilst vacuum-evaporated lms have a constant lm thickness of $330 nm. X-ray diffraction (XRD) patterns for both solution-and vacuum-processed show the main diffraction peak positions to be identical indicating that both techniques generate similar mixed-halide perovskite, Fig. 5e. The observed diffraction peaks at 14.12 , 28.44 , and 43.23 are assigned to the (110), (220), and (330) planes of the orthorhombic crystal structure. 47 The small peak at 12.65 is assigned to the (110) diffraction peak of the remaining PbI 2 compound. The best solar cell device based on the planar heterojunction architecture of FTO/c.l.-TiO 2 /CH 3 NH 3 PbI 3Àx Cl x /spiro-MeOTAD/ Ag generates solar cell parameters of J sc , V oc , FF, and PCE: 21.5 mA cm À2 , 1.07 V, 0.67, 15.4%, respectively, Fig. 5f.

Vacuum deposition of HTL
The top selective contacts (either ETL or HTL) in perovskitebased solar cells can be inuenced by the doping and environmental conditions (air, humidity, temperature, and lightsoaking) in which the cell is being operated. [106][107][108][109] Efforts have been made to nd ETL/HTL materials that are less inuenced by environmental conditions, which is expected to help minimize batch-to-batch variations.   PEDOT:PSS/polyTPD/CH 3 NH 3 PbI 3 /PCBM/Au) by varying the perovskite layer thicknesses from 200 nm to 900 nm. J sc was observed to increase as the perovskite layer thickness increased, and the rate of J sc increase was faster at the beginning up to 300 nm and slower for devices with thicker active layers. The devices with thicker perovskite layers were observed to have lower FF reducing the overall PCE. The cell with a 900 nm perovskite lm thickness was still able to generate respectable solar parameters of J sc , V oc , FF, and PCE: 19.8 mA cm À2 , 0.92 V, 0.4, 7.2%, respectively. Interestingly, the authors observed that replacing the pristine polyTPD with a slightly p-doped version of polyTPD (0.05% oxidized) in the cell with a 900 nm perovskite layer led to the signicant improvement of the FF and PCE (J sc ¼ 19.5 mA cm À2 , V oc ¼ 0.94 V, FF ¼ 0.65, and PCE ¼ 12%). This work showed that with an appropriate HTL, solar cell PCEs had only a weak dependence on the perovskite lm thickness. In addition, it demonstrated the properties of long diffusion lengths for electrons and holes in vacuum-processed perovskite lms.
Perovskite solar cells using inorganic hole conductors (such as NiO, CuI, and CuSCN) as HTLs have received attention because of their better stability than HTLs using spiro-MeO-TAD. 72,110,111 Subbiah et al. 72 reported the initial attempts of vacuum-deposited CH 3 NH 3 PbI 3Àx Cl x perovskite employing NiO and CuSCN, Table 1. Although the reported PCEs were much lower compared to those employing organic HTLs, it represents a promising step toward stability.
Schulz et al. 112 identied that V oc losses of up to 0.4 eV could arise from an ionization energy (IE) mismatch between the spiro-MeOTAD HTL (IE ¼ 5.0 eV) and CH 3 NH 3 PbI 3Àx Cl x perovskite (IE ¼ 5.4 eV). Polander et al. 74 reported fully vacuum-processed planar heterojunction CH 3 NH 3 PbI 3Àx Cl x perovskite solar cells using various p-doped HTLs with different IE values ranging from 5.0 eV to 5.6 eV and C 60 as the ETL. The authors studied the inuences of the energy level mismatch between the valence band maximum (VBM) of CH 3 NH 3 PbI 3Àx Cl x (IE ¼ 5.4 eV) perovskite and the different HTLs on the solar cell performance. It has been shown that the IE of the HTL correlates with the V oc of solar cell devices. Devices employing HTLs with IEs of up to 5.3 eV yielded a high V oc and PCE. In contrast, with IEs beyond 5.3 eV, a substantial decrease in both J sc and V oc was observed, which was attributed to the absence of driving force for hole extraction. Optimized solar cells employing spiro-MeO-TPD in a planar cell conguration of ITO/F6-TCNNQ/spiro-MeO-TPD/CH 3 NH 3 PbI 3Àx - In another study of fully vacuum-processed planar heterojunction performed by Kim et al., 73 the employment of HTL (MoO 3 /NPB) and ETL (C 60 /BCP) with a double-layer structure was observed to show improved energy level alignments at the interfacial contact resulting in higher V oc . The solar cell with ITO/MoO 3 /NPB/CH 3 NH 3 PbI 3 /C 60 /BCP/Al planar heterojunction architecture showed best solar cell parameters of J sc ¼ 18.1 mA cm À2 , V oc ¼ 1.12 V, FF ¼ 0.68, and PCE ¼ 13.7%.

Hybrid deposition method
Despite the aforementioned advantages of vacuum-based fabrication of perovskite layers and solar cells, difficulties in calibrating the QCM parameters for CH 3 NH 3 I materials were mentioned in almost all of these studies as a key challenge to achieve reproducible, uniform and stoichiometry controllable perovskite lms. 5,37,71,74,75 The evaporation rate of CH 3 NH 3 I is difficult to calibrate and control because of its relatively high vapor pressure. In addition, CH 3 NH 3 I is observed to deposit everywhere on the cold surfaces inside the chamber. For instance, the CH 3 NH 3 I layer was detected (XRD and AFM) on the top surface of a substrate that is facing the opposite direction of the CH 3 NH 3 I source. 79 In contrast, lead halides were observed to deposit mainly along the line-of-sight direction from the source. The high vapor pressure of CH 3 NH 3 I also leads to cross-talking to the reading of the QCM that is used to monitor the evaporation rate of lead halides. To solve such a challenge, Qi and coworkers developed a new methodology (the hybrid deposition method) where the perovskite stoichiometry is ensured by controlling the CH 3 NH 3 I vapor partial pressure inside the vacuum chamber. 78,79 The optimized home-built instrumentation is illustrated in Fig. 7a. 78 A more detailed study on the hybrid deposition method was reported by Wang et al. 79 and is discussed later in this section. The main vacuum chamber (Part #1 in Fig. 7a) is evacuated by using a pumping system consisting of a turbo molecular pump (HiPace 300, Pfeiffer) and a manual gate-valve (10840-CE01, VAT). The substrate holder stage (Part #3) allows stable cooling and heating in the temperature range from À190 C up to 200 C and can accommodate a wide range of substrate (Part #4) sizes up to 5 Â 5 cm 2 . A substrate shutter (Part #5) is mounted just below the substrate. The evaporation rates are monitored by two QCMs (Parts #6 and #7). The rst QCM (Part #6) facing downward monitors the PbCl 2 evaporation rate while the second QCM (Part #7) facing upward is used to monitor the CH 3 NH 3 I vapor and avoids the cross-talk from the metal halide source. Two evaporation sources are used for the sublimation of the precursor materials. CH 3 NH 3 I vapor was produced by a Knudsen cell (Part #8) type source to ll the chamber. It is emphasized that a permanent shutter in front of the Knudsen cell was mounted for avoiding the high ux of CH 3 NH 3 I reaching directly the substrate, which may cause the non-uniform composition of the lm. To achieve a high level of lm uniformity in thickness and composition as well as to provide large scale uniform evaporation (5 Â 5 cm 2 ), the PbCl 2 is resistively heated from a large dish-shaped crucible (Part #9) with $3 cm in diameter. The heating element (Part #10) consists of a tungsten wire (f ¼ 0.25 mm) wound into a spiral shape and connected to a power supply through electric feedthroughs (Part #11). The halide shutter (Part #12) allows de-convolution and extrapolation of the lead halide evaporation rate aer subtracting the CH 3 NH 3 I evaporation rate entering in the rst QCM (Part #6). The total pressure inside the chamber is monitored by using a full-range ($10 5 to 10 À7 Pa) pressure gauge (Part #13).
The initial CH 3 NH 3 I calibration and the determination of the optimized CH 3 NH 3 I : PbCl 2 ratio procedure are similar to the method described by Malinkiewicz et al. (see Section 2). 5,71 However, in the hybrid deposition, because the CH 3 NH 3 I QCM faces upwards, the QCM parameters are set to values in such a way that the signal-to-noise ratio was reasonable to monitor the CH 3 NH 3 I during evaporation. The optimized parameters were r ¼ 0.2 g cm À3 , Z-factor ¼ 0.2, and tooling factor ¼ 100. The absolute amount of CH 3 NH 3 I inside the chamber cannot be quantied. Therefore, the perovskite deposition conditions (PbCl 2 : CH 3 NH 3 I ratio) were optimized by depositing several batches of perovskite lms with varied CH 3 NH 3 I nominal rates to identify the evaporation conditions that led to strong XRD peaks measured on perovskite lms. In this way, large-area uniformity of the perovskite lms ($135 nm) was demonstrated by measuring XRD patterns at 12 different points on the 5 Â 5 cm 2 deposited lm, Fig. 7b. The hybrid-deposited lms with $50 nm and $135 nm perovskite lms were observed to show a uniform semi-transparent light-orange color with a highly reective (shiny) surface, distinctively different from the black or dark brownish color commonly observed for solution processed samples. Based on AFM measurements the surface roughness values of $4.6 nm (Fig. 7c) and $9 nm were determined for the $50 nm and $135 nm perovskite lms, respectively.
The centimeter-scale uniform semi-transparent nature of the perovskite lms grown by the hybrid deposition method is particularly suitable for large-scale window photovoltaic applications where good transparency and reasonable efficiency are prerequisites. 113,114 The best performing device for the $50 nm perovskite lm (Fig. 7d, blue curve) under standard AM1.5G illumination achieved J sc ¼ 10.5 mA cm À2 , V oc ¼ 1.06 V, FF ¼ 0.566, and PCE ¼ 6.3%. On the thicker perovskite lm ($135 nm), the measured J-V curve under illumination produced J sc , V oc , FF, and PCE of 17 mA cm À2 , 1.09 V, 0.535, and 9.9%, respectively (Fig. 7d, red curve).
Wang et al. 79 performed detailed systematic studies on the perovskite formation using the hybrid deposition method by varying the (i) evaporation source materials (PbCl 2 : CH 3 NH 3 I versus PbI 2 : CH 3 NH 3 I), (ii) substrate temperature, and (iii) postannealing conditions. The instrumentation was slightly modi-ed to position the CH 3 NH 3 I QCM facing downwards and right above the CH 3 NH 3 I evaporation source to enhance the evaporation rate detection for CH 3 NH 3 I. A shutter was placed between the CH 3 NH 3 I QCM and evaporation source. With this new geometry, the authors were able to better control the CH 3 NH 3 I rate by the QCM over a long deposition period ($1 h). This shows that the evaporation of CH 3 NH 3 I cannot be treated as standard line-of-sight evaporation (e.g., PbI 2 or PbCl 2 ) and signicant optimization in the system is needed for the better control of the CH 3 NH 3 I vapor inside the chamber during perovskite formation. Based on the PbCl 2 : CH 3 NH 3 I versus PbI 2 : CH 3 NH 3 I studies, the following reaction steps are proposed to take place for the perovskite lm formation under the vacuum conditions. PbCl 2 + 2CH 3 NH 3 I / 2CH 3 NH 3 Cl + PbI 2 (R1)

PbCl 2 + CH 3 NH 3 Cl / CH 3 NH 3 PbCl 3 (R2)
PbI 2 + CH 3 NH 3 I / CH 3 NH 3 PbI 3 (R3) In the PbCl 2 : CH 3 NH 3 I deposition case when excessive PbCl 2 is present, reactions (R1) and (R2) occur, forming a pure CH 3 NH 3 PbCl 3 phase. As the ratio of PbCl 2 : CH 3 NH 3 I reduces, the lms are composed of phase segregated CH 3 NH 3 PbCl 3 and CH 3 NH 3 PbI 3 perovskites via reactions (R1), (R2), and (R3). When the PbCl 2 : CH 3 NH 3 I ratio was further decreased matching stoichiometry, only the pure CH 3 NH 3 PbI 3 perovskite phase was observed to form and corroborated by XRD and X-ray photoelectron spectroscopy (XPS). This is also consistent with the solution-processed perovskite lms where no XPS Cl 2p signal was found in the bulk perovskite lm and only 1% Cl could be detected at the bottom 20 nm of the lm. 26 Therefore, the chemical formula of "CH 3 NH 3 PbI 3 " was more precise to be represented than "CH 3 NH 3 PbI 3Àx Cl x " in the perovskite lms formed from PbCl 2 + CH 3 NH 3 I precursors. In the PbI 2 : CH 3 -NH 3 I deposition case, only reaction (R3) takes place. The excess of PbI 2 is readily detected in XRD with a characteristic 12.6 peak, which corresponds to the PbI 2 . On the other hand, excess of CH 3 NH 3 I was observed to generate the characteristic peaks at 9.7 , 19.6 , and 29.6 . An additional peak at 11.4 was observed to evolve as a function of air exposure time and associated with the H 2 O-incorporated perovskite (complex formation). No signicant morphology differences were observed in the optimized perovskite lms deposited from PbCl 2 : CH 3 NH 3 I and PbI 2 : CH 3 NH 3 I cases showing surface roughnesses of 24.5 nm and 26.5 nm, respectively, measured by AFM. It is interesting to note that in the PbCl 2 : CH 3 NH 3 I case, stronger preferred orientation along the (110) plane of CH 3 NH 3 PbI 3 was observed to form compared to that of the PbI 2 : CH 3 NH 3 I case. It has been proposed that the additional intermediate CH 3 NH 3 Cl species formed (R1) from the PbCl 2 : CH 3 NH 3 I evaporation help slow down the reaction kinetics for the nal CH 3 NH 3 PbI 3 formation. 26,115 In the recent work by Teuscher et al., 80 a proportional-integral-derivative (PID) driven thermal evaporator was developed in their vacuum chamber allowing a more precise control of the PbI 2 : CH 3 NH 3 I stoichiometric ratio as well as improving reproducibility. The composition of the deposited materials was quantitatively analyzed using inductively coupled plasma mass spectrometry (ICP-MS) and the PbI 2 : CH 3 NH 3 I ratio of 1 : 0.96 resulted in the best performing solar cell devices (device structure: FTO/TiO 2 compact layer/ perovskite/spiro-MeOTAD/Au): J sc ¼ $18 mA cm À2 , V oc > 1.1 V, FF > 0.7, and PCE > 12%.
The growth of the perovskite lm was highly dependent on the substrate temperature during deposition, mainly inuenced by the sticking coefficient of CH 3 NH 3 I vapor. The low substrate temperature (À50 C) led to a high sticking coefficient, but a poor quality perovskite lm with partial coverage. At higher temperatures (>80 C), lms with excess PbI 2 and intermediate phases were observed and associated with the lower sticking coefficient of CH 3 NH 3 I vapor. The optimum substrate temperature was 20 C, which generated growth of perovskite lms with high crystallinity and full coverage.
Solution-processed perovskite lms generally require the post-annealing treatment (80-120 C) for efficient conversion from the precursors to perovskite and to ensure vaporization of the solvent and subsequent crystallization. 94,95 On the other hand, Malinkiewicz et al. 5,71 have shown that post-annealing is not required in vacuum-processed perovskite lms, yet attain high solar cell efficiencies. Wang et al. 79 performed detailed post-annealing studies on non-stoichiometric and stoichiometric perovskite lms formed by the hybrid-deposition method. It is shown that post-annealing is benecial for the perovskite lms with excessive CH 3 NH 3 I. Gentle annealing in an N 2 environment at 110-120 C helps desorb the undesirable H 2 O-incorporated complex. High temperatures (>130 C) decompose the CH 3 NH 3 PbI 3 to PbI 2 . The post-annealing at 120 C for 1 h in an N 2 environment on a perovskite lm with stoichiometric composition was observed to have a negligible effect on the crystallinity and morphology of the lm probed by XRD and AFM. This is in good agreement with the device performance that shows nearly the same PCE compared to that of the identically prepared cells (same deposition batch), but one with and the other without the post-annealing treatment. In this work, the solar cells based on optimized perovskite lms as thin as $170 nm generated J sc , V oc , FF, and PCE of 19.92 mA cm À2 , 1.098 V, 0.524, and 11.48%. 79 The high V oc (over 1 V) typically achieved by the hybrid deposition 78,79 is well aligned with the reported values using vacuum-deposition methods: 1.07 V in the work of Liu et al. 4 and 1.05 V in the work of Malinkiewicz et al. 5 On the other hand, solution processed ones have generally substantially lower V oc possibly due to the large variations of the lm morphology. 4 The pin-hole free uniform perovskite layer prevents shunting pathways effectively, leading to a lower recombination rate. 52,116,117 In a recent work by Zhao et al., 81 vacuum-processed perovskite solar cells with an ultrathin metal-oxide free and annealing-free C 60 or C 70 as the ETL was demonstrated to generate high PCEs (Table 1).

Hybrid chemical vapor deposition
Chemical Vapor Deposition (CVD) is a scalable technology that uses batch processing to generate throughput sufficient for industrial applications, and is used for a wide variety of material deposition. CVD is oen performed with a tube furnace, where the temperature, gas environment, and ow rate can be well controlled. Perovskite lm growth by a CVD process was rst demonstrated in 2014, 82 and can be differentiated from other vapor deposition processes that are more direction dependent and require higher vacuum conditions. CVD relies more on thermally driven diffusion, which can transport material to substrates regardless of substrate orientation. Typically perovskite growth by CVD uses an inert carrier gas such as nitrogen or argon to provide an oxygen and water free environment, but has been demonstrated using air as well.
Ambient pressure thermal diffusion driven perovskite lm growth was rst demonstrated using VASP with the PCE up to 12.1%. 62 In this system lead iodide was rst deposited onto substrates, which were then loaded into a heated (150 C), closed container with CH 3 NH 3 I. The time required for complete conversion was approximately 2 h. This is essentially a 2-step CVD process with nitrogen gas at atmospheric pressure. A similar ambient pressure perovskite growth was reported, but used a CVD tube furnace (145 C) and ambient air, which demonstrated maximum PCE of 12.2%. 84 This process also required 2 h for complete conversion.
Lower pressures increase the rate of diffusion and the rate of sublimation of organic halide, therefore low pressures allow for greater uniformity, faster deposition rates or the use of lower temperatures. Low pressure CVD growth was rst demonstrated using a 2-step process using PbCl 2 , and a two zone furnace, which demonstrated up to 11.8% PCE. 82 One zone was dedicated to CH 3 NH 3 I (185 C) and the other to substrates (145-170 C), with pressures of 100 Pa using nitrogen carrier gas. Note that perovskite lm properties strongly depend on the process temperature, and therefore precise temperature measurements close to the source or substrates inside the tube furnace can provide valuable insight for process optimization. Two zone deposition allows for faster deposition because the vapor pressure of the organic halide can be controlled independent of the substrate. The nominal reaction time reported was 1 h. A later paper by the same group using formamidinium iodide show perovskite conversion with less than 30 min of heating the organic component and an efficiency of 14.2%. 83 This suggests that the deposition times can be shortened, which is desirable for industrial processes. The typical lm roughness of perovskite samples is 20-40 nm, and the lms are semitransparent. Similar work using a low pressure 2-zone furnace was performed by different authors, but without the use of nitrogen carrier gas and at a slightly lower substrate temperature (140 C) to achieve a PCE of 12.7%. 85 However, the pressure was not specied. The highest performing reported solar cell fabricated by CVD (15.1%) used a low vacuum (100 Pa) and single zone furnace (140 C) with 2-3 h deposition time. 118 Low pressures can be used to reduce the required reaction temperature, because the sublimation temperature of organic halide is reduced at lower pressures. High performing cells were fabricated at $0.3 Pa and 82 C reaching a PCE of 14.7%. 119 However, a long reaction time of 3 h was required for complete conversion at low temperatures.
It was consistently observed that longer deposition times could be detrimental to solar cell performance, even in cases when XRD crystallinity was enhanced. 83,118 There are several possible reasons for the decrease in performance. It is possible that longer deposition times produce an excess of organic halide, which then can act as insulating contaminant in the solar cell, or can lead to more hygroscopic surfaces causing rapid decay of the cell from water absorption. In the case of chlorine-containing perovskite, longer deposition times can cause depletion of chlorine, which is believed to cause shorter carrier lifetimes and lower PCE. 83 In the case of formamidinium perovskite, the excess organic component is easily absorbed in the perovskite thin lm, creating a different crystal structure. The excess can be desorbed by annealing, but the modication of the crystal structure was observed to negatively impact the grain size and reduce performance. 83 Most processes for perovskite solar cell fabrication used a 2 step process for perovskite growth. There is a signicant difference in temperature between the evaporation temperature of metal halide and organic halide, and therefore it is difficult to uniformly deposit both layers at the same step. Most reports that deposited both layers by CVD formed discontinuous lms. This is not ideal for solar cells, but can be useful for optoelectronics devices. For instance, perovskite nano-platelets, 120,121 and nanowires 122 can be used for laser applications. Another type of single step CVD process used aerosol-assisted chemical vapor deposition for the formation of discontinuous perovskite thin lms. 123,124 One work fabricated solar cells by single step CVD and produced efficiencies up to 11.1%. 125 The advantage of using a single step approach is clear as long as uniformity is sufficient across a batch. Solar cells prepared by a CVD process typically had reasonably high stability compared to some solution processed methods. 82,83,118,119 Solution processed samples can decay even in an inert, dark environment, 50 but CVD cells kept under similar conditions were reported to be stable and even improved in efficiency over time. As the solar cell aged the V oc increased, while the J sc decreased resulting in a small net gain for the PCE. 82,83,118,119 However, it is possible that this behavior is specic to cells using spiro-MeOTAD as the hole transport medium. When kept in air, CVD solar cells were reported to decay from 14.7% to 12.1% over the course of 30 days. 119 This prolonged stability is possibly due to high temperatures during perovskite formation (less chemical and/or phase impurities) or the absence of solvent usage (e.g. DMF, DMSO) that can get incorporated into solution processed perovskite lms. However, few papers directly addressed stability under operational conditions, but it is mentioned that solar cells under continuous irradiation decay faster than cell with periodic measurements. 118

Vacuum sequential deposition
General difficulties in co-evaporation methods are oen associated with the need for careful and simultaneous control of evaporation rates of both precursors (lead halides and methylammonium halides) to achieve uniform stoichiometry in the deposited lms. In this sense, the one-material-at-a-time deposition method has a big advantage for easy control of the evaporation rates of the individual sources. 126 Chen et al. 88 reported planar structured perovskite solar cells by the sequential layerby-layer vacuum deposition method attaining a PCE as high as 15.4%, Table 1. Their devices were prepared using indium-doped tin oxide (ITO) spin-coated coated with PEDOT:PSS. Subsequently, the substrates were loaded into a high vacuum chamber (base pressure < 1 Â 10 À6 Torr) to evaporate PbCl 2 , CH 3 NH 3 I, C 60 , bathophenanthroline (Bphen), Ca, and Ag. Except for the CH 3 NH 3 I deposition, the substrate temperature was maintained at room temperature during deposition of all other layers (PbCl 2 , C 60 , Bphen, Ca, Ag), Fig. 8a. The substrate temperature while sublimating CH 3 NH 3 I was found to be critical for the photovoltaic performance, Fig. 8b and c. Photovoltaic performance was the highest when the substrate was heated to 75 C and compared to those where the substrate temperature was maintained at 65 C and 85 C during CH 3 NH 3 I deposition, Fig. 8d. The thickness of the perovskite lm was determined to be proportional to the initial thickness of the PbCl 2 layer. The lm thickness expansion ratio of $1 : 2.9 was reported when PbCl 2 is converted to CH 3 NH 3 PbI 3Àx Cl x . All the optimized lms (thickness $ 430 nm) were reported to have smooth surfaces with surface RMS roughnesses of 24.1 nm, 22.7 nm, and 23.3 nm for substrate temperatures of 65 C, 75 C, and 85 C, respectively. The smoothness of the perovskite lms for all temperatures was attributed to the typical ultra-smooth nature of the starting PbCl 2 lm (RMS roughness $7.8 nm). When CH 3 NH 3 I was deposited on the PbCl 2 coated substrate kept at room temperature, the CH 3 NH 3 I diffusion depth was limited to less than 25 nm leaving the bottom PbCl 2 layer unreacted leading to decreased PCEs. 88 Ng et al. 90 have employed the deposition of both PbI 2 and CH 3 NH 3 I at room temperature by decreasing the thickness of both precursor layers. However, an additional subsequent thermal annealing step was necessary for the full conversion to CH 3 NH 3 PbI 3 perovskite. A multilayered structure consisting of seven alternating depositions of PbI 2 (50 nm)/CH 3 NH 3 I (50 nm) pairs with subsequent annealing in N 2 gas (90 C, 1 h) generating a CH 3 NH 3 PbI 3 perovskite lm with $473 nm thickness and surface roughness of $20 nm exhibited the highest average solar cell parameters: J sc ¼ 20.0 AE 0.8 mA cm À2 , V oc ¼ 1.00 AE 0.03 V, FF ¼ 0.57 AE 0.02, and PCE ¼ 11.4 AE 0.5%. Planar structured perovskite solar cells were composed of glass/FTO/ c.l.-TiO 2 /CH 3 NH 3 PbI 3 /spiro-MeOTAD/MoO 3 /Al layers. The same group of authors investigated the impact of dry-O 2 annealing 127 of the thermally evaporated CH 3 NH 3 PbI 3 -based solar cells. The results suggested that O 2 -treatment helped enhance the solar cell performance. Under the optimized conditions, the champion device exhibited: J sc ¼ 21.8 mA cm À2 , V oc ¼ 0.96 V, FF ¼ 0.60, and PCE ¼ 12.5%. 90 Low-temperature (max. of 100 C) fabrication of hole-conductor-free planar perovskite solar cells consisting of only a CH 3 NH 3 PbI 3 /C 60 bilayer structure was reported by Hu et al. 89 to generate a PCE of 5.4%. Abbas et al. 91 have fabricated CH 3 NH 3 PbI 3 perovskites through sequential deposition of PbI 2 in vacuum and subsequently to CH 3 NH 3 I aer transferring the samples to a graphite vessel. The solar cell devices with FTO/c.l.-TiO 2 /CH 3 NH 3 PbI 3 /P3HT/Au structure and aer optimization of the P3HT layer (thickness $ 30 nm and polymer concentration $ 12 mg ml À1 ) have generated high solar cell parameters: Atomic layer deposition (ALD) can be another suitable technique for automatizing the sequential deposition of alternating layers of precursors. ALD is a low-vacuum and lowtemperature deposition technique capable of uniform, conformal growth of lms over large area with atomic thickness precision. Although, at present, there are no direct processes for the growth of OHPs fully by ALD, alternative protocols have been proposed by Sutherland et al., who showed enhanced optoelectronic properties of ALD processed OHP layers. 128,129

Flash evaporation or single-source thermal ablation technique
The concept of ash evaporation was rst described by Harris and Siegel in 1948 demonstrating the evaporation of metal alloys with controlled stoichiometry. 130 Later on, the method was applied for the evaporation of inorganic semiconductor alloys 131 as well as oxide-based perovskite materials. 132 In 1999, Mitzi et al. 97,133 demonstrated that the same technique could be applied for the fabrication of OHP lms (named as single source thermal ablation technique). Briey, the OHP material is initially synthesized by solution processing and spread on a metal heater (e.g. tantalum or molybdenum), Fig. 9a. The dried OHP and heater are loaded into a vacuum chamber. Aer the system is pumped to vacuum, a large current is passed through the heater causing the OHP material to rapidly evaporate and condense onto a substrate. The desired OHP lms are formed when the material is heated rapidly and at high enough temperatures, causing sublimation of the entire compound without thermal decomposition of the organic constituents, Fig. 9a. 93,97,[133][134][135][136] Longo et al. 93 have synthesized CH 3 NH 3 PbI 3 perovskite lms by using the ash evaporation technique and showed that smooth surface morphology was obtained with a surface RMS roughness of $17.6 nm, Fig. 9b. GIXRD measurements conrmed the formation of stoichiometric CH 3 NH 3 PbI 3 and revealed high degree of crystallinity, Fig. 9c. Solar cell devices with ITO/PEDOT:PSS/CH 3 NH 3 PbI 3 /polyTPD/ PCBM/Ba/Ag layers in a planar heterojunction structure showed respectful solar cell parameters of J sc ¼ 18 mA cm À2 , V oc ¼ 1.067 V, FF ¼ 0.68, and PCE ¼ 12.2%. 93

Fundamental understanding of OHP films prepared by vacuum processing
In parallel to several studies focusing on improving the device performance, equal effort has also been made to address the fundamental aspects of the device physics and chemistry. This The majority of the fundamental aspects of perovskites come from reports on solution-processed perovskites. 26,54,112,[137][138][139][140][141][142][143][144][145][146][147][148][149][150] Vapor-based systems used for the OHP lm synthesis offer unique advantages, i.e., they are compatible with surface science analytical tools and in situ studies. In situ monitoring allows the investigation of events taking place in the OHP lms, e.g. during formation or degradation, without altering its pristine conditions that can be inuenced by the environmental conditions (e.g. H 2 O, O 2 , temperature, light, etc.). In addition, it allows systematic investigation of the inuences of controlled environmental conditions (e.g. humidity, O 2 , temperature, light, etc.) on the material system under study. Pistor et al. 151,152 have used a dedicated vacuum chamber system where perovskite lm crystalline phase formation from the co-evaporation of PbCl 2 and CH 3 NH 3 I sources could be monitored in situ and real-time by a built-in XRD setup, Fig. 10a. The PbCl 2 : CH 3 NH 3 I ux ratio was observed as a key parameter for the formation of perovskite lms with distinct crystalline phases, Fig. 10b. The CH 3 NH 3 I (MAI) source was heated at a constant temperature of 110 C. The temperature ramp of the PbCl 2 source (T PbCl 2 ) was increased steadily from 350 C to 465 C. Under low PbCl 2 ux conditions, marked with a dotted line (1) in Fig. 10b, the formation of a dark gray/brown CH 3 NH 3 PbI 3(1Ày) Cl 3y perovskite lm with diffraction peaks at 14.046 , 28.332 , and 31.69 , phase (A), was characteristic. Additional energy dispersive X-ray spectroscopy (EDX) and optical transmission/reection measurements revealed lms with low chlorine content (y ¼ $0.02-0.05) and a bandgap of $1.6 eV. At higher PbCl 2 ux, dotted line (2) in Fig. 10b Fig. 10b. In contrast to the Br-I mixed perovskites (CH 3 NH 3 PbI 3(1Àz) Br z ) where the formation of solid solution over the whole range (0 < z < 1) was reported, 153 the authors of this study concluded that mixed CH 3 NH 3 PbI 3(1Ày) Cl y perovskites are not stable for all mixtures between CH 3 NH 3 PbI 3 and CH 3 NH 3 PbCl 3 . The authors estimated this miscibility gap to be in the range of 0.02 < y < 0.5 and explained it by the increasing difference in the I À -Br À -Cl À ionic radii and was demonstrated by Mosconi et al. 154 using rstprinciple calculations. Ng et al. 76 have also studied the formation chemistry of mixed CH 3 NH 3 PbI 3Àx Cl x perovskites by in situ XPS. PbCl 2 was evaporated layer by layer with an increasing total thickness on top of a CH 3 NH 3 I lm (15 nm) pre-deposited on ITO and XPS Pb 4f, Cl 2p, C 1s, and N 1s core levels were monitored as a function of the PbCl 2 thickness. During the initial deposition (0.2 nm to 0.5 nm) the Pb signal was detected, but no Cl signal was observed taking into account the detection limit of the XPS measurement ($AE0.1 atm%). The Cl signal is only observed for a PbCl 2 lm thickness above 1 nm indicating that PbCl 2 can be included when away from the CH 3 NH 3 I/PbCl 2 interface. The absence of Cl at the CH 3 NH 3 I/PbCl 2 interface (PbCl 2 thickness below 0.5 nm) was attributed as a result of the interaction between PbCl 2 and CH 3 NH 3 I where Cl À is energetically unfavorable because of the abrupt difference in the ionic radii of Cl À and I À ions. 154 These results are also consistent with the low content of chloride identied in solution processed CH 3 NH 3 PbI 3Àx Cl x perovskite lms. 26,155,156 The formation mechanism of pure CH 3 NH 3 PbI 3 perovskite was studied by Liu et al. 157 using in situ XPS on successive depositions of thermally evaporated CH 3 NH 3 I on a pre-formed PbI 2 lm with focus on the double C 1s feature observed, while only one N 1s species was present. The peaks observed at 286.6 eV and 402.7 eV in the BE scale were assigned to the photoelectrons originated from C and N elements in CH 3 NH 3 PbI 3 perovskite. The low-BE feature at 285.3 eV in C 1s was assigned to CH 3 I from the dissociation of CH 3 NH 3 I to CH 3 I and NH 3 . In fact, the assignment of the low-BE feature is still under debate. Ng et al. 158 and Li et al. 159,160 have also identied more than one peak in the C 1s region and assigned the low BE feature to amorphous carbon (C-C).
Li et al. 160 performed in situ XPS and investigated the degradation of co-evaporated CH 3 NH 3 PbI 3 perovskite lms under a controlled environment of dry-O 2 , and H 2 O exposures conducted in the vacuum chamber by the aid of high-precision leak-valves. CH 3 NH 3 PbI 3 was reported to be not sensitive to O 2 even at higher exposures of 10 13 Langmuir (L, 1 L ¼ 10 À6 Torr s). However, a reaction threshold of $2 Â 10 10 L was found for H 2 O exposure proposing decomposition of CH 3 NH 3 PbI 3 to NH 3 , HI, and PbI 2 in agreement with previous reports. 161,162 One important requirement for attaining high-efficiency OHP solar cells is to match the electronic energy levels of OHP absorber and the adjacent selective contacts (ETL and HTL) for minimal energy loss and reduced charge recombination. 112,142,146 A large collection of energy levels with respect to (w.r.t.) vacuum for the different materials commonly employed in OHP-based solar cells is shown in Fig. 11. 17,18,163 The indicated bandgap values for OHPs in Fig. 11 correspond to optical bandgaps (that differs from the transport gap) and all of studies were reported on solution-processed OHPs. 164 Few studies exist on the determination of energy levels on vacuum-processed perovskites measured by UPS. 157,159,160 The at band assumption widely considered can still provide a rational judgement when choosing functional layers to be coupled with the OHP layer. However, careful determination of band bending, interfacial states, and interfacial dipoles is important when considering the band alignments of OHP layers with adjacent functional layers. 112,142,146,147 In this sense, lm deposition by vacuummethods is suitable for studying the energy alignments because step-wise deposition with controlled incremental amounts of lm thickness can be conducted as well as it is directly compatible with UPS and IPES systems. Energy level diagrams at interfaces were determined for the following material systems: CH 3 CN). 142 Hysteresis appears as a manifestation in the J-V curves, where its shape is strongly inuenced by the scan direction (forward versus reverse), scan rate, scan history, pre-illumination, and pre-biasing conditions. [165][166][167][168][169][170] It is pointed out that hysteresis in OHP-based solar cells will lead to inaccurate reporting of PCEs, leading to inaccuracies when comparing the different reported PCEs among different laboratories. Stabilized power output under working conditions has been suggested as a useful parameter. 165 Efforts have been made on determining the origins of the hysteresis phenomena. Recently, several studies have enforced to ion migration as the main phenomena in determining charge transport in OHP materials. 8,[170][171][172][173][174][175][176] Few studies reported on the hysteresis-free J-V characteristics in vapor-processed OHP solar cells. 71,85,88,91,93,177 However, its origin is difficult to be understood from the reported studies as systematic studies are lacking.

Conclusions and outlook
The identication of new properties in novel materials is key for generating new technologies. The introduction of well-established inorganic or organic materials led to dramatic improvements in the history of technology. Inorganic silicon-based technology enabled unprecedented development of advanced electronic devices (e.g. laptops, smart phones, digital cameras, Si solar cells, etc.). Organic semiconductors on the other hand have received attention in light-weight, exible, at-paneldisplay and organic light emitting diode applications. The past few years have witnessed a rapid evolution of hybrid organicinorganic OHP-based solar cells. OHP materials effectively combine the properties of the inorganic framework and the intercalated organic species. As has been reviewed previously, OHP materials show several desirable properties for photovoltaic [178][179][180] and other optoelectronic technologies: 181-183 high absorption coefficients, long carrier diffusion lengths, ambipolar carrier transport, shallow defect levels, emission efficiency, and low concentration of defects. Although the operation of vapor based methods oen require a certain level of vacuum that is generally perceived as high cost processing, they have been widely employed in the semiconductor industry demonstrating high throughput and reliability. Therefore, a detailed factual cost analysis would be required for evaluating various fabrication methods taking into account the costeffectiveness in a mass production scenario. [184][185][186] When transferring solar cell technology from laboratory scale fabrication know-how to industry-scale production, fabrication cost, efficiency, and lifetime are the three major factors, which are associated together with functionality (e.g. transparency, exibility, easy integration in tandem cells, etc.). 185,187,188 At this stage, despite the superior quality of lms prepared by the vapor-deposition method (e.g. uniform and full coverage), 4 best efficiencies achieved for OHP solar cells based on vapor-based methods (a PCE of 16.5% with an active area of 0.2 cm 2 ), 37 (Table 1) are still somewhat lower than that of the solution-processed cells (a certied PCE of 20.1%). The efficiencies of vapor deposition based perovskite solar cells are still lagging behind their counterpart prepared by solution based methods. On the other hand, we regard this more as a strong motivation to invite more research effort on vapor based methods, rather than an intrinsic disadvantage associated with vapor based methods. Since the ground-breaking work by the groups of Prof. Snaith and Prof. Bolink in 2013 on vapor deposition prepared perovskite solar cells, more and more research groups became interested and have been making contributions to this topic. For example, at this stage it has been demonstrated by a number of groups that vapor based methods provide another viable route to fabricate perovskite solar cells leading to promising efficiencies using both regular and inverted structures (see Table 1 and Fig. 12). Efficiencies as high as 16.5% have been reported so far and the trend of efficiency increase is very clear as shown in Fig. 12. Further improvements on the performance of vapor deposition based perovskite solar cells may come from the following strategies. First of all, multiple reports have indicated that the properties of vapor deposition based perovskite are distinctively different from those prepared by solution processing. Therefore, it is necessary to carry out more in-depth investigations on vapor deposition based perovskite lms. Secondly, due to the property difference, the optimal fabrication conditions for vapor deposition based perovskite solar cells are most likely not the same as those for solution prepared ones. A complete set of optimization processes hopefully will provide further efficiency growth of vapor deposition based perovskite solar cells. Thirdly, vapor deposition based perovskite lms have some specic advantages, e.g. high degree of uniformity (even at relatively thin lm thicknesses) and semitransparency, which may provide the vapor based methods with unique features beyond the pure efficiency considerations. Note that none of the PCE values reported in Table 1 has been certied, which to some degree also underscores the need to further explore these vapor-based methods. In general, caution should be practiced when comparing the cell performance reported by different laboratories as PCE values in OHP solar cells are shown to strongly depend on measurement conditions, such as the voltage-scan polarity (forward versus reverse) and rate, light-soaking, prebiasing, and cell temperature. 168,189 OHP lms deposited by vapor-based methods generally show low XRD intensity, but the corresponding solar cells still show relatively high efficiencies. No direct correlation has been reported between XRD peak intensity (generally considered as a measure of crystallinity) and the corresponding solar cell efficiency. In addition, the crystallinity, perovskite composition, and lm morphology are strongly dependent on the choice of the substrate (e.g. FTO, TiO 2 , SiO 2 /Si, etc.), which is expected to become critical when designing tandem cell architecture: the optimized evaporation conditions may differ according to the substrate on which OHP lms are being deposited. With regard to vacuum systems that are usually needed for vapor based processing, special care is required for pumps (especially turbo molecular pumps).
For commercialization to take off, large area modules will be required. 190,191 A few attempts were made on large area (1 cm 2 ) fabrication based on vacuum-and CVD processes, Table 1, showing a promising PCE as high as 13.84%. 92 Because uniform, high-quality, and full-coverage OHP lms are achievable, vacuum-and CVD processes are expected to have unique advantages in the fabrication of large-area OHP solar modules. 190 The OHP materials are hygroscopic in nature, and are susceptible to degradation upon the intake of moisture. 192 Therefore, for protecting the core material in the OHP-based solar cells it requires stringent encapsulation. In addition to extrinsic degradation factors by moisture, the intrinsic stability of perovskites remains a major issue. The chemical reactions between moisture and perovskite need be carefully studied to Fig. 12 Progress of solar cell efficiency in vapor-processed perovskite solar cells. The graph was generated based on the reported efficiencies shown in Table 1. A trend of increase in the number of published works on vapor-processed perovskites is inferred from the graph. unravel the reaction pathways, which provide insight for the development of methods (e.g. chemical composition engineering 2,153,193 ) for stabilizing perovskites. The negative standard Gibbs free energy for iodide perovskite degradation was also reported in the absence of moisture. 161,192,194 The majority of stability tests reported in the literature, provides the lifetime proles of OHP solar cells under the storage conditions (in N 2 or in ambient air). Only limited data exist for stability proles under real operating conditions for vacuum-or CVD processed OHP solar cells. 109 A major drawback of high efficient OHP solar cells is the use of Pb 2+ , the material toxicity of which has been emphasized. 184,190,195 Efforts have been made to nd alternatives such as Sn 2+ . However, the instability of Sn 2+ to form Sn 4+ leads to a metal-like behavior and lowers the photovoltaic performance. 196 Many other elements in the periodic table (e.g. Co 2+ , Fe 2+ , Mn 2+ , Pd 2+ , and Ge 2+ ) were suggested as alternatives for Pb 2+ . 24,163 Through comparison with the amount of lead used in lead acid batteries, a much lesser amount of lead is estimated to be required to produce 1000 GW per year from CH 3 NH 3 PbI 3 perovskite solar cells. 197 Therefore, tracking and minimizing the amounts of lead salt as well as quantifying solar cell efficiency normalized by the perovskite amount (e.g. PCE/thickness parameter calculated in Table 1) are proposed as important parameters to evaluate the toxicity at the times of disposal, recycling, and of eventual accidents. 184,190,195