Research Update: Strategies for improving the stability of perovskite solar cells

The power-conversion e ffi ciency of perovskite solar cells has soared up to 22.1% earlier this year. Within merely five years, the perovskite solar cell can now compete on e ffi ciency with inorganic thin-film technologies, making it the most promising of the new, emerging photovoltaic solar cell technologies. The next grand challenge is now the aspect of stability. The hydrophilicity and volatility of the organic methylammonium makes the work-horse material methylammonium lead iodide vulnerable to degradation through humidity and heat. Additionally, ultraviolet radiation and oxygen constitute stressors which can deteriorate the device performance. There are two fundamental strategies to increasing the device stability: developing protective layers around the vulnerable perovskite absorber and developing a more resilient perovskite absorber. The most important reports in literature are summarized and analyzed here, letting us conclude that any long-term stability, on par with that of inorganic thin-film technologies, is only possible with a more resilient perovskite incorporated in a highly protective device design. C 2016 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons license [http: After merely four years since its realization as solid-state device, 1 the perovskite solar cell has achieved an e ffi ciency of 22.1% exceeding even that of multicrystalline silicon. 2 That in itself is remarkable for a technology as new young as the perovskite cell. However, more impressive is the fact that the perovskite

After merely four years since its realization as solid-state device, 1 the perovskite solar cell has achieved an efficiency of 22.1% exceeding even that of multicrystalline silicon. 2 That in itself is remarkable for a technology as new young as the perovskite cell. However, more impressive is the fact that the perovskite absorber can be processed simply from solution and/or vapor. [3][4][5] The fact that the efficiency is now on par with that of established photovoltaic technologies spurs the dreams of a rapid transition from a laboratory technology to an industrial-scale solar power technology. Without the complex material processing of its competitors and the abundance of its fundamental materials, the perovskite solar cell could thus provide even cheaper clean energy.
Simple processing comes at a price for the perovskite solar cell. Because the material is soluble in aprotic polar solvents, it degrades readily and quickly when it comes in contact with even small amounts of such solvents such as water. 6 The big challenge for the perovskite solar cell, which it needs to overcome in order to become a mature photovoltaic technology, is the aspect of stability so that it may have a chance to comply with the common standard for photovoltaic modules to retain their most of their initial efficiency for up to 25 years. 7 In this Research Update, we will give a short and concise overview over the current efforts to enhance the stability and durability of perovskite solar cells. We can differentiate between two fundamental strategies: the first one is focused on the entire device seeking to guard and protect the absorber from external assaults by developing specialized functional barrier structures. The second strategy seeks to improve the resilience and stability of the absorber itself. This can be done by either altering the elemental composition of the perovskite or by modifying the perovskite absorber with functional molecules with the purpose of making the perovskite less susceptible to, for example, moisture degradation. Various excellent reviews (see Seigo Ito et al. in this issue) give an extensive overview over the current understanding of the weaknesses of the most common perovskite absorbers and the most detrimental external forces. 6,[8][9][10] We will therefore confine ourselves to a concise descriptive account of the most common degradation pathways in order to give context for the strategies to halt and prevent them.
In the very first attempts to integrate the perovskite absorber into photovoltaic devices, methylammonium lead triiodide (MAPbI 3 ) was used as sensitizer in liquid-state dye-sensitized solar cells. 11,12 The authors noted that the perovskite would degrade within minutes when in contact with acetonitrile and ethyl acetate, which were used as electrolyte, respectively. This observation already pointed towards the inherent vulnerability of the perovskite material to aprotic polar solvents. The most ubiquitous member of this class of solvents is water. In fact, the structural integrity of perovskite films was shown to be compromised by the presence of atmospheric moisture within a timeframe of a few hours. 13 The degradation mechanism has been documented and described in detail. [14][15][16] Essentially, because the organic methylammonium cation is only weakly bound via hydrogen bonds to the ionic cage of lead and iodide, a water molecule can break these bonds. As a consequence, the methylammonium is solubilized by water and afforded greater mobility to move throughout the structure. During the formation process of the perovskite, this can be beneficial by facilitating the removal of excess methylammonium. 17 Similarly, a limited exposure to moisture after complete crystallization can improve the optoelectronic properties of the perovskite. 17 However, the increased mobility of the solubilized methylammonium also destabilizes the material because external forces such as heat and light can lead to the removal of the volatile organic constituent, which results in the decomposition to lead iodide. 13 In the presence of external forcing, this decomposition process can be initiated by individual molecules of water. 16,18 This shows that the methylammonium based perovskite is highly sensitive to even minuscule amounts of water. An integral part of devices employing this perovskite absorber must therefore be a barrier to prevent moisture ingress.
Following the initial incarnation of the solid-state perovskite solar cell, the most common device architecture is still the n-i-p structure, in which electrons are collected by the transparent electrode at the bottom; while the top-layer is a hole-collecting layer extracting the holes. Being the top layer of the device, the hole-transporter can be considered the first line of defense against moisture ingress.
A variety of approaches has been explored to improve the moisture-blocking properties of this layer. A focal point of many studies has been the relatively poor barrier property of the most commonly employed hole-transporting materials, namely, spiro-OMeTAD, poly(triarylamine) (PTAA), and poly(3-hexylthiophene) (P3HT), to effectively prevent moisture ingress. 13 The combination of heat, light, and moisture generally results in rapid degradation of the perovskite layer underneath the hole-transporters, which occurs even more rapidly when the hole-transporting materials (HTMs) are doped with a hygroscopic dopant such as Li-TFSI. 13,14,19,20 Removing the need for such a dopant is an obvious strategy to improve the moisture resilience and overall device stability. [21][22][23][24][25][26][27] This can be achieved by replacing the reactive dopants with inert additives such as carbon nanotubes or graphene, 21,28 or by pre-oxidizing a fraction of the hole-transporter in order to improve the charge transport characteristics of the undoped hole-transporting material. 23,29 Nguyen et al. could show that by pre-oxidizing a small fraction of spiro-OMeTAD, the overall conductivity of the hole-transporter is increased significantly, yielding device performances comparable to devices with conventional Li-TFSI doping but more stable performance under illumination. 23 Leijtens et al. showed that the same technique of partial pre-oxidation can be used on a range of other small-molecule hole-transporters to improve their charge transport characteristics. 29 Removing the need for external dopants can also be achieved by synthesizing novel hole-transport materials which have tailored energy levels and inherently good charge transport characteristics and therefore do not require additional doping for efficient charge extraction. 22,24,25,27,30 Promising results were obtained by Kim et al., employing a newly synthesized polymer as hole-transporter which was shown to protect the perovskite absorber for 1400 h at 75% relative humidity in the dark and at room temperature ( Fig. 1(c)). 27 Unfortunately, the authors compare their dopant-free polymer to doped spiro-OMeTAD on the basis of hysteretic current-voltage scans, failing to report the steady-state performance. 27 To further boost the resilience against moisture ingress into the device, hydrophobicity has become an additional important aspect informing the synthesis of new hole-transporting materials. In fact, hole transporters specifically designed to have intrinsic hydrophobic properties are able to halt creeping moisture ingress and thus improve the device stability. 13,19,22,29,31,32 In our study, we developed a two-layer system composed of single-walled carbon nanotubes for selective charge extraction and an encapsulating polymer matrix such as poly(methyl methacrylate) or polycarbonate to protect the device from moisture ingress. 13 The devices remained stable in ambient humidity conditions at elevated temperatures of 80 • C, while control devices were shown to rapidly degrade (Figs. 1(a) and 1(b)). 13 The protective effect of the encapsulating polymer even allowed the direct exposure to running water for a short amount of time.
Leijtens et al. demonstrated that their hydrophobic hole-transporter can be resilient under even harsher stability tests such as full water immersion. 29 Moisture penetration of the hole-transporting layer can also be achieved by crosslinking the hole-transporter to thusly minimize the moisture ingress pathways. 26 Xu and co-workers synthesized an arylamine derivative (N4,N4 ′ -Di(naphthalen-1-yl)-N4,N4 ′ -bis(4-vinylphenyl)biphenyl-4,4 ′ -diamine) (VNBP) which can be thermally crosslinked. 26 A thin layer of MoO 3 deposited on top p-dopes the polymer without compromising its barrier function, and allowing it to achieve a steady-state efficiency of 16.5%. 26 Even more impressively, the VNBP-MoO 3 double layer protects the perovskite absorber from degradation and retains its high performance under harsh stressing conditions. Devices were shown to withstand thermal stressing at 110 • C for 1 h ( Fig. 1(d)) and high humidity exposure at 70% relative humidity for 30 days. This resilience against degradation can probably be credited to the excellent barrier properties of the crosslinked double-layer hole-transporting layer preventing moisture ingress and methylammonium egress. This is illustrated by the fact that devices remained unaffected even when being exposed to both polar and non-polar solvents. 26 The resilience against solvents may in particular hold a lot of promise for tandem applications.
One of the most compelling long-term testing regimes has been reported by Abate et al. who were investigating silolothiophene-linked triphenylamines as hole-transporters. 33 Using maximum power-point tracking, the steady-state performance of devices with one of the doped triphenylamine hole-transporters and doped spiro-OMeTAD was continuously measured under ultraviolet (UV)-filtered simulated sunlight in an argon atmosphere at around 45 • C for up to 1000 h. 33 By fitting a double-exponential decay to the power-output of the devices, the authors could extract a half-life for the two compared materials. The triphenylamine outperformed spiro-OMeTAD with a half-life of 6000 h, compared to 1000 h, which the authors attribute to a better thermal stability. 33 Further studies of this sort will be very insightful with respect to the device stability, in particular, if they were to include relevant stressors such as moisture, oxygen, and UV light. A range of inorganic hole-transporter materials were proposed with the aim of improving the device stability, including copper thiocyanate (CuSCN), 34-41 copper iodide (CuI), 42,43 and nickel oxide (NiO x ). 44 The latter for example showed good stability over 60 days, during which control devices with spiro-OMeTAD underwent rapid degradation. 44 Evidently, next to high power-conversion efficiencies and a simple synthesis, the aspect of stability has become an important benchmark for new hole-transporter systems. 45 In numerous publications, beneficial properties of hole-transporter have been claimed; however, a majority of those studies relies on "cupboard aging" or "shelf-live assessment," where devices are stored in a dark and often dry environment, and tested at different time points. It is worth pointing out that this kind of "aging" is insufficient to assess the overall stability of functional devices, in general, and the protective effect of the hole-transporter, in particular, because the driving forces of the absorber degradation have largely been removed. 6 In order to provide unassailable evidence for improved device stability, devices should remain operational for an extended period of time being exposed to either full solar spectrum illumination, an increased temperature around 85 • C, an external load equivalent to maximum power-point operation, or a combination thereof. Stability is clearly moving more and more into the focus of the development of new holetransporters for perovskite solar cell; however, to-date, the highest efficiencies are still obtained using the established hole-transporter materials spiro-OMeTAD and PTAA. 5,46,47 Both these holetransporters have been shown to lack the ability to effectively block moisture ingress. 13 However, there is another strategy that seeks to enhance the device stability against moisture for these materials, by introducing a very thin layer of an insulating water-impermeable material between the perovskite absorber and the hole-transporter. Several such barrier materials have been explored and shown to reduce moisture ingress and improve stability in a non-dry environment. [48][49][50][51][52] Using atomic layer deposition (ALD) as a technique to deposit an ultrathin layer of Al 2 O 3 between the perovskite and spiro-OMeTAD, the stability of a device kept at RH 50% without illumination for 24 days could be improved and its performance was reported to still be at 90% of the initial value. 48 Guarnera et al. showed that a simply spin-coated layer of alumina nanoparticles at the interface of the perovskite absorber and spiro-OMeTAD improves performance and stability. 49 A sealed device with the buffer layer experienced a performance decrease of merely 5% over 350 h of continuous full-spectral illumination at AM 1.5. The performance of a control device without the buffer dropped to around 40% of its initial efficiency within the same time frame. 49 Several hydrophobic organic layers have been employed at the interface between the MAPbI 3 perovskite and spiro-OMeTAD to act as moisture barriers. The layers have to be thin enough to allow charge transfer but thick enough to effectively block moisture ingress. Pentafluorobenzenethiol, for example, was shown to increase the device durability at a relative humidity of 45%. 53 Modifying the perovskite surface with dodecyltrimethoxysilane (C 12 -silane), the devices lost merely 15% of their initial efficiency after 600 h under ambient conditions. 54 The main objective of many barrier layers is to hamper any moisture diffusion through the hole-transporter to prevent degradation. However, several studies have shown that another source for device impairment is chemical reactions between the iodide ions from the perovskite and the metal electrode, typically Ag or Al. [55][56][57][58] This can result in the formation of an insulating barrier such as Ag- which will impair charge extraction at this electrode. 59 Long-term operability of these devices requires therefore, that, additionally to moisture ingress from the outside, the diffusion of ions from the perovskite to the electrode is blocked in order to avoid electrode corrosion. Sanehira and co-workers found that this degradation pathway can be blocked for aluminum by inserting a thin layer of MoO x between the spiro-OMeTAD layer and the metal electrode. 50 Back et al. approached this issue by focusing on chemically stabilizing the ionic defects thought to be responsible for the iodization of the electrode by introducing a chemical inhibition layer at the interface with the metal electrode ( Fig. 2(b)). 55 Greater electrode stability could also be achieved by Guerrero et al. by moving away from using either Ag or Al, and instead employing a Cr 2 O 3 /Cr electrode which is shown to be chemically inert towards iodide. 58 Similarly, Au electrodes do not undergo any corrosion; however, the price of the metal is prohibitively high for large-scale commercialization. Ku et al. have pioneered a very promising approach of employing a thick carbon electrode. 60 The carbon cathode is inexpensive, is not prone to corrosion, and, interestingly, can act as hydrophobic moisture barrier additionally increasing the stability of a perovskite solar cell. [60][61][62][63][64] In some architectures, devices using the carbon cathode do not even require a selective p-type contact. 60 triple-layer structure composed of mesoporous TiO 2 and mesoporous ZrO 2 , which is then infiltrated with MAPbI 3 and contacted by a thick carbon layer. 61 In the initial report, stable performance under full sunlight illumination, unencapsulated in ambient air for a period of more than 1000 h was reported. 61 The stability of this architecture has been further demonstrated since with a range of stress-tests including outdoor testing (Figs. 2(c) and 2(d)). 66 An encapsulated device was reported to show a remarkably stable performance when aged outdoors for seven consecutive days in Jeddah, Saudi Arabia. Further, the performance was also reported to remain stable when an encapsulated device was thermally stress-tested by keeping it 80-85 • C for 90 days, in the dark. Photostability of the perovskite devices was also demonstrated by continuously monitoring the power-output of an unencapsulated device in an argon atmosphere at 45 • C and at maximum power-point tracking conditions. However, not the full sunlight spectrum was used in this experiment, instead an light-emitting diode (LED) array served as light source, emitting only in the visible range without including contributions from the UV region. In light of previous studies observing photo-induced degradation of perovskite devices due to the photoactivity of mesoporous TiO 2 for the UV range, 67 a more rigorous photostability study with a full AM 1.5 spectrum matching the spectral range of natural sunlight is therefore needed in order to obtain a complete picture of the "real-world" stability of this architecture. The device performances of p-i-n perovskite solar cells using CIL/Ag and Ag electrodes, as a function of the J-V sweep operation time under N 2 ( Fig. 2(b)). The inset also represents the device performance, as a function of MPP tracking time under N 2 .
The discussed strategies to improve the device stability thus far have been mainly targeted at the inherent vulnerability of MAPbI 3 to moisture and the subsequent degradation of the absorber. As discussed before, being a highly polar solvent, water has the ability to solubilize the organic constituent of the perovskite, thusly compromising the structural integrity and stability of the material, in particular, in the presence of exterior stressors. The stability of perovskite devices can therefore be significantly improved by preventing or minimizing the ingress of moisture.
However, there is mounting evidence that oxygen also induces rapid perovskite degradation. 6,67-71 Ambient oxygen in combination with light can lead to direct photo-oxidation of the MAPbI 3 perovskite. 68,69 Those studies suggest that the accumulation of photogenerated electrons accelerates the oxygen-induced degradation of the perovskite absorber. Possible strategies for mitigating the oxygen assault on devices could therefore include effective barrier layers and oxygenimpermeable encapsulants, as well as efficient electron extraction layers. Several studies have reported improved device stability when mesoporous layers of TiO 2 particles or nanorods are employed as rapid electron extraction pathways. 70,[72][73][74] This strategy has to take into account, however, that metal-oxides such as TiO 2 are themselves photoactive materials for UV-induced redox reactions. [75][76][77] In their study, Leijtens and co-workers investigated the impact of UV-irradiation on perovskite devices on mesoporous TiO 2 , observing that photoinduced oxygen desorption over time generates shallow traps below the conduction band edge thus resulting in decreased performance. 67 C 60 was used by Wojciechowski as interface modification layer for TiO 2 and as stand-alone electron-accepting layer. 78,79 The latter was shown to yield a more stable device performance under full-spectrum illumination for 500 h, whereas the control device with TiO 2 as n-type layer experienced a significant reduction in performance. 79 Ito et al. employed a buffer layer of Sb 2 S 3 between TiO 2 and the perovskite, thus significantly improving the device stability under light exposure. 34 The protective effect of Sb 2 S 3 is credited to the ability of the inorganic layer to block the photocatalytic decomposition of the perovskite at the interface with TiO 2 , which is a well-known photocatalyst. 76,77 The control devices, in which the perovskite was coated directly onto TiO 2 , were shown to degrade rapidly which the authors attribute to a UV-induced photo-oxidation process at the TiO 2 surface. 34 This vulnerability to UV-activated degradation processes at the TiO 2 interface was confirmed by Li et al. 80 In this study, the authors were able to stabilize the perovskite absorber by inserting a thin layer of CsBr between TiO 2 and the perovskite, thus inhibiting the photocatalytic decomposition of the absorber. 80 Another strategy to impede performance losses due to photoactivated absorber degradation was demonstrated by Pathak et al., who doped the mesoporous TiO 2 layer with aluminum. 81 Devices with the Al-doped TiO 2 layer exhibited a much enhanced operational stability under full illumination in an inert atmosphere. The increase in stability was attributed to the substitutional incorporation of Al in the anatase lattice, thus passivating UV-generated oxygen vacancies acting as electronic trap sites on the TiO 2 surface. 81 While Al-doping makes the devices more stable, it also causes the conduction band edge of TiO 2 to shift upwards, which negatively impacts the charge extraction efficiency and results in a reduced photocurrent. This drawback can be overcome by using neodymium (Nd) instead of Al as dopant, leaving the conduction band position of TiO 2 unchanged. Devices with a Nd-doped mesoporous TiO 2 layer were shown to reach a steady-state efficiency of up to 18.2% while exhibiting a significantly enhanced stability compared to their undoped counterparts. 82 Hwang and co-workers went one step further by completely replacing TiO 2 with CdS as the n-type layer in order to improve the device photostability. 83 The authors report that after continuous illumination for 12 h, devices with TiO 2 as n-type layer experienced a significant decrease in efficiency from 15.4% to 3.1% whereas the devices with CdS had retained more than 90% of their initial efficiency. 83 The devices discussed thus far had the same n-i-p architecture, in which the device stack is configured such that electrons are extracted at the bottom by an n-type layer, and the transparent electrode, whereas holes are extracted at the top by a p-type layer and an opaque electrode. In this configuration, moisture ingress through the p-type conductor and photo-oxidation on the n-type interface cause degradation.
Conversely, this architecture can also be inverted to form a p-i-n structure. The n-type contact is typically formed by phenyl-C 61 -butyric acid methyl ester (PCBM) while the most commonly used transparent hole-selective contact is poly(3,4-ethylenedioxythiophene) poly(styrenesulphonate) (PEDOT:PSS). 84,85 The biggest stability concern is the acidic and hygroscopic nature of PE-DOT:PSS which may lead to degradation of the perovskite absorber at the interface. 86 To prevent direct contact between the corrosive PEDOT:PSS and the transparent electrode, thin inorganic blocking layers were proposed. 87,88 Such bilayer structures of PEDOT:PSS and the buffer layer were reported to experience less of a performance decrease than pristine PEDOT:PSS. 87,88 The thin interlayer of CuAlO 2 and MoO 3 are credited, respectively, with suppressing corrosion of the transparent indium-tin oxide (ITO) electrode by the acidic PEDOT:PSS. 87,88 In a similar vein, to mitigate this electrode corrosion, GeO 2 was blended into the PEDOT:PSS dispersion prior to its deposition. The reasoning was that GeO 2 due to its alkaline nature in aqueous solutions may be able to buffer the acidity of PEDOT:PSS to mitigate any ITO corrosion. In devices the hybrid p-type layer showed a much more stable performance than pristine PEDOT:PSS but still dropped to mere 50% of the initial performance within 26 h. 89 Thus far stabilizing devices with PEDOT:PSS as hole-extraction layer appears to be quite difficult with PEDOT:PSS attacking both of its interfaces. Choi and co-authors therefore abandoned PEDOT:PSS altogether and achieved a significant improvement in device stability in air by employing a pH-neutral polyelectrolyte. 86 Reduced graphene-oxide may also be a candidate as an inert and stable replacement for PEDOT:PSS. 90 However, while devices with RGO appear to be less prone to degradation when exposed to ambient air-as opposed to devices with PEDOT:PSS-the overall performance still decreased significantly within 140 h. 90 A material which holds a lot of promise as a stable hole-transport layer for devices with the p-i-n architecture is the inorganic NiO x , which is known for its good environmental stability. Compared to PEDOT:PSS devices with copper-doped NiO x exhibit in fact a significantly increased air stability, retaining around 90% of their initial efficiency after dark ageing for 240 h in ambient air, whereas PEDOT:PSS devices experienced a rapid decrease of more than 50% within 144 h. 91 Following up with a solution-processed approach to depositing NiO x , Kim et al. showed this inorganic p-type layer can also be applied on flexible substrates yielding more air-stable flexible devices. 92 In the inverted architecture, the top layer of the device is a thin n-type layer, typically PCBM. 84,93 With a thickness of merely tens of nanometers, it constitutes a poor barrier against water and oxygen penetration. In their study, You et al. therefore tackled the issue of instability for both charge-selective layers, namely, PEDOT:PSS and the PCBM, by replacing the two organic layers with inorganic ones, NiO x on the p-side and ZnO on the n-side. 94 Kept in ambient air in the dark, the TABLE I. Overview of literature values on long-term stability strategies, including relevant parameters such as relative humidity, atmosphere, illumination, and UV irradiation. The figures of merit are the initial power-conversion efficiency, the rate of decreasing efficiency (in %/100 h), the total decrease in efficiency, as well as the decline in efficiency of the respective control system.

Reference
System Structure  device with the inorganic charge-transport layers retained most of its initial performance for up to 60 days. 94 The outermost modification to enhance the stability of a perovskite solar cell is the encapsulation of a device. This strategy involves a material which is resilient against moisture and oxygen permeation, and which is used to fully encapsulate the device with strategic placements of the electrode strips which can still be accessed without compromising the integrity of the protective encapsulation. Much work on the effectiveness of various sealing techniques has been carried out in the field of organic solar cells, as oxygen-induced degradation is one of the major factors compromising the stability of organic solar cells. The most effective method is the encapsulation of the device with a glass or metal plate using a slow-permeation epoxy material as sealant. [95][96][97] This technique should be perfectly adequate for perovskite devices as long as they are fabricated on rigid substrates. Hwang et al. showed that they could also improve the performance lifetime of a device by spinning a layer of amorphous Teflon on top of the device. 98 Chang et al. use the concept of a dense alumina layer as protective barrier and show that the atomic layer deposition of an Al 2 O 3 layer significantly improves the air-stability due to the very low oxygen and water vapor transmission of the barrier layer. 99 More challenging is the device encapsulation for devices on flexible substrates. Weerasinghe and co-workers show that full encapsulation is needed for a flexible device. 100 Employing a commercial plastic barrier film, stable performance can be achieved over a period of up to 500 h. In Table I, an overview is given of stability metrics reported in literature for various approaches to stabilize CH 3 NH 3 PbI 3 . 100 To date, the inherent instability of MAPbX 3 materials towards moisture, heat, and ultraviolet (UV) light in presence of oxygen has been reported. 6,69 As discussed, methylammonium-based PSC tends to hydrolyze in the presence of moisture, thus triggering a non-reversible degradation process. 52,101,102 Although it has been reported that intermediate monohydrate phases can revert back to MAPbI 3 , further degradation results in a final decomposition products of PbI 2 and aqueous CH 3 NH 3 I, which further decompose into volatile compounds such as CH 3 NH 2 , HI, or I 2 . [14][15][16]52,103,104 Similarly, HC(NH 2 ) 2 PbI 3 decomposes to HC(NH 2 ) 2 I and HI, where HC(NH 2 ) 2 I further decomposes to the volatile sym-triazine and NH 4 I. 105,106 Moreover, it has been reported that MAPbI 3 also structurally degrades in an inert N 2 atmosphere when bulk powders and single crystals heated at temperature in excess of 200 • C, 107,108 while thin film can show signs of degradation at temperatures of 85 • C. 109 Here, we discuss recent advances in improving the inherent instability of the photoactive absorber material used in photovoltaic devices by tuning the perovskite composition.
A proposed route for stability improvement focused on the tuning of the ABX 3 general structure, where A is a cation, B is a divalent metal ion, and X is a halide. Seok and co-workers have reported on an improved stability to moisture when by controlling the halide composition. 110 By partially replacing the iodide with a bromide anion for the MAPbX 3 perovskite system, the authors were able to form a cubic three-dimensional (3D) perovskite structure phase of the Pm3m space group, instead tetragonal I4/mcm space group. This phase transition results in a slight rotation of the PbX 6 octahedrons, thus reducing the octahedra tilting and lattice distortion. 110 Tuning of the X halide composition, in the general ABX 3 formula, can also be achieved via incorporation of an entire functional group, notably thiocyanate (SCN) group. Due to its similar ionic radius of 0.215-0.220 nm for SCN − compared to 0.22nm for I − , this pseudohalogen can replace a "true halide" such as the I − halogen ion. 111,112 The partial halide substitution, forming a MAPbI 3−x (SCN) x perovskite structure, have had a beneficial impact on grain sizes and trap density, 111 optoelectronic properties, 113 and most importantly on its material stability. [114][115][116] In the case of MAPb(SCN) 2 I, the interaction between Pb 2+ and SCN − is much stronger than in the case of neat MAPbI 3 . 114 This stronger interaction suppresses the initial degradation mechanism involving the formation of hydrated intermediate containing isolated PbI 4− 6 octahedral. 14 Jiang et al. reported that MAPb(SCN) 2 I decomposed at a significantly slower rate than MAPbI 3 when exposed to a 95% RH atmosphere. 114 Furthermore, a later study demonstrated that the MAPbI 3−x (SCN) x perovskite, fabricated via a two-step sequential deposition in a humid atmosphere using a Pb(SCN) 2 precursor instead of the conventional PbI 2 route, produced higher solar cell performance. 116 The  MAPbI 3−x (SCN) x solar cells without encapsulation retained 86.7% of the initial average efficiency, when stored in a 70% RH atmosphere for over 500 h, while MAPbI 3 lost nearly 40% of its original efficiency for an identical aging condition. Interestingly, it has also reported been reported that Pb(SCN) 2 can also be used to form a two-dimensional (2D) perovskite material: (CH 3 NH 3 ) 2 Pb(SCN) 2 I 2 . 115,117 Two-dimensional structures with generic structural formula (A) 2 (CH 3 NH 3 ) n−1 M n X 3n+1 , where n is an integer, have been reported to have improved the perovskite's moisture stability due to its hydrophobic nature (Figure 3(a)). 118,119 By partially replacing the small MA + cation with a long-chain organic cations such as n-butylammine (BA) or C 6 H 5 (CH 2 ) 2 NH 3 (PEA), the perovskite can form in a multilayered compound. 118,119 However, these homologous 2D materials have less ideal properties for solar cell applications due to a wider optical band gap. This may limit their use for tandem solar cell or luminescent applications.
A recently proposed route for stability improvement focuses on the tuning of the methylammonium A cation. Cation mixtures have been reported in the early stages of the emergence of PCS; for example, Pellet et al. introduced FA into the MAPbI 3 to extend the absorption onset, while Jeon et al. incorporated MAPbBr 3 into the FAPbI 3 structure to enhance its crystallinity and structural stability, thus improving its power conversion efficiency. 120,121 However recently, cation mixtures have gained attention as a means to increase the absorber material's resistance to moisture and heat degradation. These stability gains were achieved by substituting MA with formamidinium (FA), 122 cesium (Cs) cations, [123][124][125][126] or a mixture of the two (Figure 3(b) and 3(d)). 5,127-131 Eperon et al. reported that degradation was considerably slowed for FAPbI 3 compared to MAPbI 3 when heated at 150 • C. 122 Moreover, MAPbI 3 perovskite undergoes a reversible phase transition from tetragonal to cubic when heated above ∼54-57 • C. 107 This change in symmetry, occurring within the operating temperature range of the solar cell, can potentially lead to an accelerated degradation of methylammonium-based cells when operated in real-life operating conditions. On the other hand, Lee et al. report that FAPbI 3 shows no phase transition in the range from 25 • C to 150 • C, thus eliminating significant lattice shrinkage during standard operating temperature range. 132 Although the MAPbI 3 perovskite has received significant research attention, the PSC field may ultimately choose to move away from this perovskite structure due to its inherently thermally unstable nature.
In spite of several advantages of FAPbI 3 over MAPbI 3 , including having a narrower band gap and higher thermal stability, FAPbI 3 PSC have had limited success. This is partly due to the structural instability of the black trigonal (P3m1) perovskite polymorph in the presence of moisture, resulting in a phase transition to a yellow hexagonal non-perovskite (P63mc) polymorph. 133 Stabilization of the black perovskite phase was achieved by substituting FAI for MAI or MABr, resulting in FA 1−x MA x PbI 3 or FA 1−x MA x PbI 3−y Br y perovskite compositions. 120,121 According to Binek et al. the stabilization of the 3D black FA-based perovskite, achieved with the incorporation of a smaller MA cation compared to FA, can be attributed to the larger dipole moment of MA, which arranges the PbI 6 octahedra with a pseudocubic symmetry via I-H hydrogen bonding or the increase in Coulomb interactions within the structure. 131 As a result, films remained stable and did not undergo any phase transition in the 25-250 • C temperature range. 131 However, the relatively volatile nature of the MA cation may still remain problematic for long-term stability.
Several research groups have recently reported on the incorporation of an inorganic A-site cation in the perovskite structure, notably cesium. Fully inorganic cesium-based perovskite has gained interest due to their potential of withstanding high temperatures. [123][124][125][126] CsPbI 3 forms in a cubic (pm3m) perovskite structure with a band gap of 1.73 eV. 122,123 Unfortunately, CsPbI 3 is highly unstable in this crystal structure when exposed to air and quickly undergoes a reversible phase transition to a yellow orthorhombic phase material, unsuitable for solar cell applications. On the other hand, CsPbBr 3 is less susceptible to moisture compared to their iodide counterpart. 125,134 Although they show comparable power-conversion efficiency (PCE) to that obtained by MAPbBr 3 , its wide optical bandgap of ∼2.25-2.36 eV is not ideal for single-junction solar cell applications. 125,135 Several groups have suggested a compromise between neat iodide and neat bromide perovskite by employing a CsPbI 2 Br composition. 124,126,136 This mixed-halide composition offers a greater resistance to moisture than the neat iodide composition, while utilizing a narrower band gap than the neat bromide perovskite (Figure 3(e)).
Although purely inorganic CsPbX 3 perovskites have exhibited outstanding thermal stability, to date, their reported efficiencies have yet to compete with their organic counterpart. Nevertheless, it has been shown that the cesium cation can also be used to stabilize the FAPbI 3 crystal structure, instead of relying on relatively volatile MA cation. 128,129 Lee et al. reported that cesium not only increases the PCE via stabilization of black FAPbI 3 perovskite phase but also improves film stability against light and humidity. Pristine un-encapsulated FAPbI 3 devices showed 81% degradation after 30 min of white light illumination (100 mW cm −2 ) in an ambient condition (RH < 40%), whereas FA 0.9 Cs 0.1 PbI 3 devices degraded by 67% for an identical time period (Figure 3(c)). 128 A subsequent study found that the shelf-life stability of FA 0.85 Cs 0.15 PbI 3 solar cells was greatly extended compared to neat FAPbI 3 . The FA 0.85 Cs 0.15 PbI 3 devices showed no loss in their PCE when stored in the dark for 15 days in a 15% RH atmosphere. 129 Further reports also included the use of a halide mixture in combination with the FA/Cs cation mixture, resulting in improvements in crystallinity of the material. 127,130 Moreover, reports of adding a small amount of Cs to the MA 0.17 FA 0.83 Pb(I 0.83 Br 0.17 ) 3 structure, forming a "triple-cation," have shown enhancements in stability, reproducibility, and PCE. 5 Although the MA cation remained in the precursor solution, the triple-cation composition was able to maintain 90% of its original PCE, from 20% to 18% PCE, after 250 h of aging under solar illumination using a maximum power point tracking (MPPT), in an N 2 atmosphere.
Perovskite solar cells have achieved efficiencies which are already on par with inorganic thin-film technologies; the focus will therefore need to shift towards improving their long-term stability. The inherent instability of MAPbI 3 towards moisture and oxygen can be combatted by introducing functional barrier layers into the device structures, which allow efficient charge extraction while minimizing the ingress of degradation agents. This has been shown to improve device stability to some degree in the short run. But in many cases, this appears to delay rather than to prevent the absorber degradation. The second strategy stabilizes the perovskite absorber itself by substituting its constituent ions. Mixed cation and mixed halide compositions, as well as lower-dimensional structures, show very promising results. In essence, the weak point of the perovskite is the A-site cation, of which methylammonium is simply too volatile to be retained in the perovskite structure when exposed to external stressors; in contrast, the two alternatives, formamidinium and cesium, give the perovskite much better stability, but can only be stabilized in the right phase when present as ionic mixtures. While these optimized compositions improve the stability of the perovskite, they will still require highly resilient charge-selective layers, either inorganic or inert organic layers, and mature encapsulation techniques in order to survive true long-term stability studies and systematic stress tests such as high-humidity and full-temperature cycles, which will be an important gauge for the maturity of this photovoltaic technology vis-à-vis its inorganic thin-film competitors.