REVIEW ARTICLE | APRIL 24 2023 Monitoring the stability and degradation mechanisms of perovskite solar cells by in situ and operando characterization Fanny Baumann ; Sonia R. Raga  ; Mónica Lira-Cantú  APL Energy 1, 011501 (2023) https://doi.org/10.1063/5.0145199 CrossMark   View Export Online Citation 08 November 2023 15:51:28 APL Energy REVIEW scitation.org/journal/ape Monitoring the stability and degradation mechanisms of perovskite solar cells by in situ and operando characterization Cite as: APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 Submitted: 3 February 2023 • Accepted: 29 March 2023 • Published Online: 24 April 2023 Fanny Baumann, Sonia R. Raga,a) and Mónica Lira-Cantúa) AFFILIATIONS Nanostructured Materials for Photovoltaic Energy Group (NMPE), Catalan Institute of Nanoscience and Nanotechnology (ICN2), Nanostructured Materials for Photovoltaic Energy Group, Autonomous University of Barcelona (UAB), Bellaterra, Barcelona 08193, Spain a) Authors to whom correspondence should be addressed: sonia.ruizraga@icn2.cat and monica.lira@icn2.cat ABSTRACT Solar energy technologies are among the most promising renewable energy sources. The massive growth of global solar generating capac- 08 November 2023 15:51:28 ity to multi-terawatt scale is now a requirement to mitigate climate change. Perovskite solar cells (PSCs) are one of the most efficient and cost-effective photovoltaic (PV) technologies with efficiencies reaching the 26% mark. They have attracted substantial interest due to their light-harvesting capacity combined with a low cost of manufacturing. However, unsolved questions of perovskite stability are still a concern, challenging the potential of widespread commercialization. Thus, it is imperative to advance in the understanding of the degradation mech- anism of PSCs under in situ and operando conditions where variable and unpredictable stressors intervene, in parallel or sequentially, on the device stability. This review aims to debate the advantages behind in situ and operando characterization to complement stability-testing of PV parameters in the strive to achieve competitive stability and reproducibility in PSCs. We consider the impact of applying single and multi-stressors under constant monitoring of alterations observed in PSC components or complete devices. We outline key future research directions to achieve the long-term stability necessary for the successful commercialization of this promising PV technology. © 2023 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/5.0145199 I. INTRODUCTION in perovskite stability have important implications for real-world Photovoltaic (PV) technologies are important and competitive energy yields, challenging the potential of widespread commer- means to produce electricity from sunlight, offering the potential cialization. Factors that can induce degradation to metal halide to transform our society into a low-carbon economy. However, perovskite (MOHP) materials and devices, such as humidity, atmo- for the technology to become competitive, the leveled cost of elec- sphere, bias voltage, temperature, or light exposure, have been the tricity (LCOE) must be reduced, and the best strategy is to boost center of multiple studies and debates for many years. The ISOS power-conversion efficiency and extend its operational lifetime. stability assessment protocols, elaborated after a consensus among However, the efficiency of solar cells is limited by the so-called leading international laboratories, have been recently upgraded to Shockley–Queisser (SQ) limit.1 With silicon cells (SiSC) showing incorporate stressors involving the peculiarities of PSCs. Figure 1 current power-conversion efficiency (PCE) values of 26.7% and depicts a schematic representation of the existing ISOS protocols halide perovskite solar cells (PSCs) at >26%,2,3 these two powerful developed originally for organic solar cells (OPVs) and recently PV technologies are almost at the limit of their maximum theo- upgraded for PSC (in rectangles).5,6 retical SQ value. This indicates that further significant advances in Lifetimes of 10 000 h have been reported for PSCs under dark PV efficiency practically require breaking the SQ limit and one pos- conditions (ISOS-D protocol),7 constant light irradiation (ISOS-L sible method to achieve this is by the use of multijunction solar protocol),8 and under various relative humidity (RH) conditions cells where theoretical calculations indicate PCEs around 45% if tan- (ISOS-RT),9 also calculated 30-year lifespan under outdoor opera- dem configurations are employed.4 Unfortunately, unsolved issues tion conditions can be found in the literature (ISOS-O protocol).10,11 APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-1 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape FIG. 1. Schematic representation of the main ISOS protocols employed for OPV and PSC stability assessment: ISOS-L, light irradiation; ISOS-D, dark conditions; ISOS-LT, light/humidity thermal cycling; ISOS-T, thermal cycling; and ISOS-O, outdoor testing. Updated PSC protocols (in rectangles): ISOS-LC, light/dark cycling; ISOS-I, intrinsic stability; and ISOS-V, applied voltage.5,6 08 November 2023 15:51:28 However, these works are scarce, and they do not always fulfill the of a PSC under stress factors that are well described and known to requirements of the application of stability stressors under real and impact exclusively PSCs (as described in the ISOS protocols), we will operando conditions; thus, the stability of PSCs still represents enor- be able to compare results among laboratories and understand and mous limitations. The understanding of the different degradation construct a long-lasting and enduring PSC technology. mechanisms taking place in materials, solar cells and modules, will We will also unveil the significance of in situ characteri- only be effective if the applied stressors are as close as possible to zation, which is validated by the reversibility of effects and the an actual functioning device. Outdoor analysis (ISOS-O protocol), short timescales of recovery observed in PSCs. This differs from where multiple stressors are imposed on the sample with great vari- post-mortem and ex situ characterization that can draw inaccurate ability and doses, is one of the most accurate methods to evaluate conclusions regarding mechanisms occurring during PSC failure PSC stability. At the laboratory stage, in situ and operando character- and their long-term prevention. In addition, in situ characterization ization techniques are excellent options to evaluate changes observed carries out under multi-stressor also permits to monitor synergetic, in the solar cells under single or multiple stressors with time. competitive effects, with the aim of gaining greater understanding of In this review, we present a comprehensive and critical sum- the gaps regarding the elusive stability properties of metal-organic mary of the state-of-the-art knowledge regarding in situ and halide perovskite (MOHP) in all varieties of device structures. operando characterization during the process of degradation of PSCs This review is organized based on the degree of the different under introduced stressors. We also give insight into the capabilities, stressors applied to the PSCs (Fig. 2). First, we summarize research the accuracy, and the few limitations of in situ characterization tech- works employing single stressors such as humidity and oxygen, light niques. Our aim is to dive deeper into relevant research works to gain irradiation, thermal stress, bias-loading, and cycling stressors. We insight into the different degradation mechanisms observed in PSCs. finalize the review with the summary of those works where two or We also seek to persuade the reader of the advantages of in situ and more stressors are employed. operando characterization under multiple stressors, as these require- ments are closer to the real working conditions of PSC. The reader II. HUMIDITY AND OXYGEN will also notice our incessant association of the reviewed works to External environmental conditions, such as moisture or oxy- ISOS protocols (Fig. 1). The ISOS protocols are internationally rec- gen, suppose a challenge for the long-term stability of most mate- ognized experimental conditions that should be followed for the rials; therefore, the risks of corrosion and oxidation should always stability assessment of PSCs. These have been identified under a be considered. Photovoltaic components of PSC, particularly the consensus among international experts in the field. The applica- MOHPs, are highly sensitive. tion of ISOS protocols is necessary because most stability studies are As perovskite is exposed to moisture, water molecules pene- conducted under laboratory conditions without following standards trate the polycrystalline film crystal lattice, mainly through the gaps or protocols and focusing mostly on one main stress factor.12,13 and surface defects at the grain boundaries. If or when reaching However, only by continuously monitoring the degradation sufficient concentrations, reactions occur, either in a reversible or APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-2 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 2. Degradation in PSCs. (1–4) Schematic representation of a PSC depicting the different stressors affecting an in operando PSC: (1) illumination, (2) heat, (3) electrical bias, and (4) atmosphere (moisture and oxygen). (a)–(e) A representation of some of the most characteristic degradation mechanisms observed in halide perovskites: (a) halide/phase segregation, (b) formation of hydrates, crystal decomposition by (c) releasing volatile molecules methylamine (MA), I2 , NH3 , and HI or (d) forming precipitates (Pb0 , I2 , etc.), and (e) morphology changes (e.g., smaller grain size). irreversible pathway. Water in the lattice form hydrates via hydro- be done by compositional tuning (cation or halide),17,18 adding gen bonding,14 i.e., the perovskite crystal structure is separated into passivating organic molecules(e.g., binding with dangling bonds of smaller units, such as two-dimensional structures. Hydration results surface atoms)19 or creating a physical barrier that hinders water in the loss of all or part of the optoelectronic properties of the MOHP penetration.20 layer. Water incorporation is reversible by drying,15 healing the sep- In this article, we will review and put in perspective the recent arated units back to perovskite structure. However, when new grain debate about the pathways in which humidity affects PSC at vari- boundaries are formed or when reaction products evaporate (HI, ous conditions, which could be studied in detail, thanks to in situ CH3 NH2 ) or precipitate (Pb0 , I2 ), changes in morphology or chem- characterization setups. Although largely solved by encapsulation, ical reactions are permanent. Relative humidity exposure, transport air-degradation was one of the first issues faced by the PSC research at the interface of the crystal, and crystal integrity are determining community, we will outline how in situ characterization has been factors for the outcome; formation of reversible monohydrate, irre- used to bring clarity into the mechanisms of physical and chemical versible dehydration, permanent chemical reactions, or drying. In degradation under controlled humidity and O2 stressors. contrast, mixed cation perovskite compositions undergo more com- In the year 2015, two works opened an important discussion plex degradation pathways that can involve phase separation.16 As about perovskite hydration. Yang et al. performed both in situ we conclude in Sec. III A (light + ambient) and Sec. VII (multistres- ultraviolet-visible optical absorbance (UV–vis) and in situ grazing sor), the sole presence of O2 has no serious effects on perovskite incidence XRD (GIXRD) measurements on MAPbI3 to study degra- stability unless an external input of energy is added to promote dation kinetics.21 They observed a dependence of relative humidity reactions. Oxidation of small molecules and polymers employed as (RH) on the absorbance reduction rate at 410 nm. The degrada- selective contacts in the PSC architecture can however be a source tion “half time” evolved from 4 to 34 h, and from 1000 to 10 of instability. This modifies charge transport and energy alignment 000 h upon the reduction of RH from 98% to 80% and from 50% properties (e.g., in the oxidation of spiro-OMeTAD), thus affect- to 20%, respectively, while in dry air no changes on the UV–vis ing device performance via the conductivity of collecting layers or spectra happened for 2 weeks. With in situ GIXRD, real-time mon- the carrier mobility at interfaces. For the MOHP layer itself, most itoring of 2D scattering showed that additional diffraction signals essential functions remain intact upon exposure to oxygen in the appeared at q = 5 and 7 nm−1 upon exposure to 80% RH. These dark. If MOHP bonds are strengthened to enhance the integrity new peaks were hypothesized to be an intermediate hydrate phase of the 3D perovskite structure, air stability is improved. This can that could revert to the perovskite structure. Indeed, the absorption APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-3 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape spectra showed partial recovery of the absorption edge at 760 nm observed under pure N2 or O2 , but, upon adding moisture, a after flushing with dry nitrogen. That same year, the formation of decrease in resistance over time was observed that could be recov- perovskite hydrates was revealed by combining time-resolved XRD ered back by exposing samples to dry N2 , but long-time exposure led and in situ ellipsometry. Leguy et al. explored the crystal changes to PbI2 formation. The results fully support chemisorption followed during hydration (80% RH) and drying (35% RH) of 270 nm thin by hydration in the formation of a monohydrate phase. Song et al. film MAPbI3 on glass.15 They concluded that an initial 6% expan- used laser beam-induced current (LBIC) imaging to create time- sion of the crystal lattice occurred by the interpenetration of water dependent maps of the external quantum efficiency (EQE) at a fixed molecules into the structure, leading to the creation of monohy- wavelength. While scanning a 532 nm laser beam across the sample drates in the entire perovskite grain [Fig. 3(1b)], while massive every 6 min and measuring the spatially resolved photo-generated expansion to 28% of the initial MAPbI3 volume was attributed to current, at N2 flow with ∼50% RH, four stages of degradation the formation of dihydrate crystals on grain surfaces Fig. 3(1c). [Fig. 3(2a)] were observed:26 Stage 1: EQE increased until reach- Additionally, changes in the refractive index acquired by in situ ing a peak around 10–15 min after condition application. Stage 2 ellipsometry during hydration/dehydration revealed that the macro- (18–105 min): EQE steadily decreased from ∼72% to ∼56%. Stage 3 scopic conversion to monohydrate is isotropic. Similar results were (2 h): a front of sharp EQE decrease swept over the device result- extracted in a more recent work by using GIXRD on full solar cell ing in merely ∼6% EQE. These three stages were reversible, followed devices at 60% RH.22 Results from Yang and Leguy were temporar- by an irreversible fourth stage [Fig. 3(2b)] with loss of the remain- ily considered conflicting regarding the initial hydrate product,15,21 der of the EQE signal at ∼1300 min. All devices tested showcased and detailed studies on the electrical properties and degradation the same general behavior [Fig. 3(2c)]. The authors identified dop- mechanisms of perovskite films followed. ing of spiro-OMeTAD and the perovskite interfaces as the cause for To confirm the results available and further detail the mecha- stage 1–2, reversible perovskite-hydration as stage 3, and irreversible nism of degradation under humidity conditions, Li et al. conducted hydration as stage 4.26 detailed studies on MAPbI3 films in p-i configuration (without elec- In mixed perovskite compositions, contrary to initial obser- tron transport layer (ETL) and top electrode) using in situ scanning vations on MAPbI3 perovskite, the mechanisms of ambient degra- force microscopy (SFM).23 Their results supported the mechanisms dation significantly differ. Hu et al. performed in situ XRD on outlined by Leguy et al. but found irreversibility in the morphol- four different perovskite films on glass, with various A-site cation ogy changes occurring simultaneously with MAPbI3 hydration. The combinations (MA, FA, Cs, and Rb) and different halide (I/Br) ratios.17 They demonstrated that Rb ions in the lattice enhanced 08 November 2023 15:51:28 initial smooth crystallites in the film started changing to steps and grooves after 4 h under 80% RH during time-resolved morphology moisture instability. When films were exposed to 75% RH air, room studies with SFM. Complementary in situ XRD spectra taken every T and dark for 60 min using a hydration-chamber made from 15 min (80% RH N2 ) revealed the formation of a monohydrate peak x-ray-transparent polymers, FAMA and FAMACs (Cs: 5 mol. %)- (MAPbI3 ⋅ H2 O) starting at 5.5 h and increasing intensity over time, based films showed minor signs of degradation, only a small PbI2 coinciding with the observed morphology changes from SFM. After peak appeared after 60 min. Rb-containing films, however, showed 9 h, dry N2 gas flow conditions were applied in both experiments to the onset of a new peak at 11.4○ after only 15 min, followed by explore the reversibility. While the changes of the XRD spectra were three additional reflections at 22.9○ , 34.7○ , and 46.8○ . The peak at reversed, the morphology changes were not; i.e., the newly formed 2θ = 11.4○ was attributed to non-photoactive phases identified as grain boundaries persisted.23 With in situ grazing incidence neutron RbPb2 I4 Br and affinity for side-phases containing Rb and Br to ther- scattering (GISANS), Schlipf et al. could see that much more water modynamic favoritism toward lower Goldschmidt tolerance-factor is adsorbed on MAPbI3 crystal surfaces and accumulated in amor- in-between these ions. Two complex and concomitant degradation phous regions than the amount perceived in the crystal hydrates mechanisms for Cs0.05 (MA0.17 FA0.83 )0.95 Pb(Br0.17 I0.83 )3 (CsMAFA) previously.24 By using moist air with D2 O molecules, they could mixed perovskite with in situ XRD [Fig. 3(3a)] under 85% RH were determine that, at low RH, D2 O is not incorporated into the crys- found by Kazemi et al.18,19 One pathway involved the decompo- tal structure but rather located in amorphous regions around the sition of the initial structure through dissolution/recrystallization, crystallite grains. These amorphous regions can be visualized with leading to PbI2 . A second pathway, in which phase segregation hap- GISANS, whereas only crystalline degradation products can be visu- pens, results in the formation of the identified phases (Cs-poor and alized using XRD. At ∼73% RH, a volume expansion is attributed I-rich)-CsMAFA perovskite and CsPb2 Br5 [Fig. 3(3b)]. Similar to to the formation of the monohydrate, and, at >93% RH, dihydrate the above, Urwick et al. compared different mix MOHP with vary- phases are formed inside the MAPbI3 structure. Below 73% RH, the ing A-site cations (MA, MAFA, CsFA, and CsMAFA) using in situ trend of water uptake is found to exhibit a plateau between 30% time of flight (ToF) GISANS [Fig. 3(4)] under 90% RH up to 12 h.27 and 60% RH attributed to the saturation of crystal surfaces with ToF-GISANS enabled the identification of low atomic weight phases adsorbed water. Smaller crystallites are more affected both regarding with high depth resolution. They concluded that Cs increased the D2 O ingress and hydration, making degradation strongly dependent film stability, preventing the formation of lead halide sub-products. on morphology. It was observed that the bulk endured a more drastic decay than the Since many works focused on the perovskite material and film surface, suggesting that degradation initiated in the bulk. Both surface science to identify the ambient-induced degradation, the these reports hinted at crystal integrity and eventually segregation of real-time effects on the optoelectronic response of the perovskite mixed cation perovskites. were less explored. Hu et al. used in situ electrical resistance mea- Since air exposure is external and can be isolated via pro- surements of perovskite films with buried interdigitated electrodes tection, studies have focused on implementing ambient-protection and with a controlled atmosphere.25 No change of resistance was strategies for PSC, and some works employed in situ measurements APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-4 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 3. (1) Structural configurations of (1a) MAPbI3 cubic phase, (1b) monohydrate, CH3 NH3 PbI3 ⋅ H2 O, and (1c) dihydrate, (CH3 NH3 )4 PbI6 ⋅ 2H2 O. Reprinted with permis- sion from Leguy et al., Chem. Mater. 27, 3397 (2015). Copyright 2015 American Chemical Society. (2a) LBIC EQE maps (at 532 nm) of a PSC after exposure to 50% ± 5% RH. (2b) high-resolution color scale LBIC images for stage 4 (480–1320 min). (2c) Areal average LBIC EQE (at 532 nm) as a function of time after exposure to humidity. Reprinted with permission from Song et al., Adv. Energy Mater. 6(19), 1600846 (2016). Copyright 2016 John Wiley & Sons. (3a) In situ x-ray diffraction patterns in contour plot of CsMAFA film as a function of aging time under 85% RH. (3b) Evolution of the x-ray diffraction peak intensity as a function of aging time. Reprinted with permission from Kazemi et al., Energy Environ. Mater. 6, e12335 (2023). Copyright 2023 John Wiley & Sons. (4) GISANS experiment setup for in situ moisture degradation. Reprinted with permission from Urwick et al., Energy Rep. 8, 23–33 (2022). Copyright 2022 Elsevier. to validate their hypothesis. Yang et al. included a study on the Kazemi et al. demonstrated that modifying the perovskite com- protective effects of various HTL, showing that while PTAA and position or passivating it with an ammonium additive decreased P3HT acted as barriers slowing the rate of decomposition, spiro- air-induced degradation substantially.18 Parallel in situ XRD and OMeTAD accelerated the degradation. Kundu and Kelly performed in situ liquid cell TEM on perovskite films exposed to 85% RH in situ absorbance spectroscopy at saturated humidity (99% RH) air flow demonstrated the known behavior of MAPbI3 hydration flow to prove that a tailor-made hole transport material (HTM) described earlier. Emphasizing that the preferential growth of can work well as a protective barrier against humidity.20 They pre- both initial tetragonal MAPbI3 and PbI2 degradation product was pared a poly(methyl metacrylate) (PMMA) film with embedded in the [110] direction while intermediate degradation structures P3HT nanowires HTL layer to act as a H2 O-barrier protector. In situ grew preferentially in the (001) plane-direction, they supported absorbance spectroscopy at 410 nm revealed that PMMA protec- that hydration originates primarily through crystal planes with tion was much greater than P3HT. Decomposition of unprotected more affinity to incorporate water molecules. In contrast, when MAPbI3 happened in under 6 h, MAPbI3 with P3HT degraded 9 mol. % excess of PbI2 was present in the MAPbI3 film, the in less than 24 h, and pure PMMA covered MAPbI3 only slightly degradation mechanism skipped the hydrated intermediate. The degraded after 100 h. Decomposition was sequential; when pro- onset of PbI2 formation was delayed, but once initiated, degra- tected samples started to degrade, pinholes formed in the bar- dation proceeded faster, and further, including PbIOH formation. rier followed by rapid degradation of the perovskite underneath. A passivation approach using tetrapropyl ammonium ion (TPA+ ), APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-5 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape (TPAx MA(1−x) PbI3 ), showed drastic improvement of stability, with- III. ILLUMINATION out the formation of hydrates and minimal PbI2 formation after Illumination is an unavoidable part of achieving photovoltaic 400 min. The authors proposed that TPA+ molecules passivate response in solar cells. Light photoexcite carriers in the perovskite surface defects, particularly unbounded lead, hindering H2 O pene- provide the desired electric current, but the free electrons can also tration into the crystalline structure. Maniyarasu et al. incorporated trigger unwanted chemical reactions that are detrimental to the sta- 1-octyl-3-methylimidazolium chloride ionic liquid (IL) into the bility. Illumination effects also seem to be the strongest when carriers FA0.9 Cs0.1 PbI3 perovskite to stabilize it and tracked the humidity- are accumulated inside the structure without being collected (i.e., in induced degradation with near ambient pressure x-ray photoelec- films without contacts or devices at an open circuit). tron spectroscopy (NAP-XPS) from 0% RH to 30% RH.28 It was concluded that cation-mixture improved moisture stability but that the IL additive did not further improve the performance in humid A. Extrinsic material reactions conditions. The pristine IL-perovskite showed signs of water incor- While perovskite film is rather insensitive to pure oxygen in the poration as an O1s core level signal that increased under humidity dark,29,30 the photoexcited electrons under illumination react with conditions, indicating continuous incorporation of water in the IL. oxygen molecules creating superoxides (O2 − ) that further catalyze a Other peaks corresponding to perovskite did not undergo many rapid dissociation of the crystal.31 Similar reactions can happen with changes, in contrast to other studies done on MAPbI3 . The incorpo- the H2 O molecules in moist air. Whereas almost saturated air with ration of water disrupts the passivation of the ionic liquid, yet as we relative humidity (RH) close to 95% can lead to perovskite decompo- will mention in Sec. IV, the additive showed usefulness for thermal sition in the dark, lower RH does not induce perovskite degradation stability.28 unless it is triggered by illumination. With the above examples, ongoing debate about the detrimen- The decomposition pathways are diverse, depending on the tal effects of humidity on perovskite, more specifically regarding stress conditions and the perovskite stoichiometry. Below we pro- the pathways that result in PbI2 degradation product, has to some vide a review of the most relevant in situ studies that addressed the degree been solved. On the basis of complementary works, the mois- synergetic effects of light and ambient stressors. ture ingress in MAPbI3 perovskite causes a reversible intermediate Senocrate et al. studied the MAPbI3 degradation kinetics by monohydrate that evolves into a permanent perovskite dihydrate.15 in situ XRD and UV–vis absorbance of the film under O2 gas flow Thanks to the in situ studies, the kinetics of the formation of with and without illumination.30 The perovskite gradually depleted 08 November 2023 15:51:28 these hydrates are known for different humidity conditions. Large- in both O2 and light until only PbI2 absorption was left, the PbI2 grain thin films are less receptive to humidity-induced degradation [0 0 1] peak dominated the XRD spectra, and the decay rate because of the reduced area for moisture ingress, i.e., water molecule increased proportionally with light intensity. Adding a small partial permeation is enhanced at surface defects and gaps between crys- pressure of I2 [P(I 2 ) = 1.4 ⋅ 10−5 bars] substantially decreased the O2 - tallites. That vapor-induced degradation occurs via initial deproto- induced degradation, reinforcing that I2 is one of the sub-products nation of the MA+ by water to form MA (gas) and hydrated HI, of such degradation. Since air contains both oxygen and water, other apart from the final PbI2 , has consequently been disputed. Such works employed air with varying humidity conditions to be closer to highly volatile products would be flushed out if created, leading to real conditions. irreversible degradation. When reversibility was seen even in open Tang et al. showed unchanged MAPbI3 2θ peaks for 24 h in chambers and during gas-flow,23,26 it was proof of intermediate steps dark and 40% RH, which faded almost immediately upon illumi- before decomposition. This provides hope that after-treatments, nation [Figs. 4(1c) and 4(1d)].32 In a vacuum, the major degrada- such as drying, can remedy early degradation before proceeding to tion was attributed to MA loss, which accelerated under illumina- chemical decomposition. Isolation of devices from atmosphere gases tion vanishing MAPbI3 signal within 6 h [Figs. 4(1a) and 4(1b)], by encapsulation and passivation are very effective, therefore some- leading to crystalline PbI2 and Pb0 [Fig. 4(1b)]. Nakamura et al. what rendering humidity discussions extinct. Yet, production and reported more detailed MAPbI3 degradation products using 2D accumulation of degradation products can be self-catalytic; small wide-angle x-ray scattering (2D-WAXS) [Fig. 4(2a)].33 Under simul- errors in fabrication where humidity remains can be detrimental to taneous light (100 mW/cm2 ) and humidity (95% RH) exposure, long-term stability. For PSC technology to fulfill its low-cost fabri- MAPbI3 degraded into smaller crystallite sizes and crystalline forms cation promises, stability against oxygen and humidity must be fully of NH4 PbI3 ⋅ 2H2 O and CH3 NH3 I (but no PbI2 ). Dark conditions understood and controlled. in 95% RH only formed PbI2 and increased MAPbI3 grain sizes A lack of consensus in the literature regarding dry and [Figs. 4(2c) and 4(2d)], in contrast with the previous report in which humid air is observed, while in some works 30% RH is referred 40% RH in dark did not trigger any XRD changes.32 Probably there is to as “drying” conditions,24 other works use UHV or pure O2 a threshold on the humidity concentration that is detrimental for the or N2 gas,28 while 30% RH is considered “humid.” Our pro- perovskite. Ho et al. reported a different degradation mechanism on posed general reading among these discrepancies is that, if water FA0.85 Cs0.15 PbI3 .34 Coupled atomic-force microscopy (AFM) with molecules are dissociated from the crystal, it is “drying.” We photo-thermal induced resonance (PTIR) enabled (<100 nm) spa- would like to clarify that because of single stressing conditions, tial resolution mapping of the FA cation [Fig. 4(3)]. The mixed some authors argued that only the moisture component in ambi- cation perovskite segregates into δo -CsPbI3 needle crystals and ent conditions plays a role in perovskite degradation, while O2 large δ-FAPbI3 grains that further degrade to PbI2 . In this work, itself seldom had any effect. As will be presented later, this is 10 mW/cm2 LED illumination and 20%–85% RH were applied greatly disputed, once other stressors are added O2 concentration is ex situ in the AFM chamber with the instrument off to avoid laser important. damage. By using Raman spectroscopy, Pistor et al. monitored APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-6 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 4. (1) In situ XRD of CH3 NH3 PbI3 films at 350 K under (1a) vacuum/dark, (1b) vacuum/light, and under air without (1c) or with (1d) light illumination. Reproduced with permission from Tang et al., J. Mater. Chem. A 4, 15896 (2016). Copyright 2016 The Royal Society of Chemistry. (2a) System for in situ 2D-WAXS measurement under light radiation and humid conditions. Time evolution of integrated 1D-WAXS patterns of CH3 NH3 PbI3 polycrystalline thin films under (2b) illumination and humid conditions and (2c) humid conditions. Adapted with permission from Nakamura et al., ACS Mater. Lett. 4, 2409–2414 (2022). Copyright 2022 American Chemical Society. (3) Schematic showing AFM and PTIR images obtained on the same area. AFM morphology (orange) and PTIR (blue) at 1712 cm−1 under high humidity and light. Below, the schematic shows formamidinium iodide (FAI) evaporates, and depressions form. Adapted with permission from Ho et al., ACS Energy Lett. 6(3), 934–940 (2021). Copyright 2021 American Chemical Society. MAPbI3 film surface in air and found PbI2 formation after 5 min of for the migration of ions, and induces strain in the crystal lattice. laser irradiation (260 W/cm2 ).35 Several examples found in the literature refer to this phenomenon with different terms: “halide segregation,” “phase separation,” “ion B. Intrinsic material reactions migration/distribution,” or “ion accumulation at interfaces/grain Incident photons under illumination excite carriers in the boundaries.” The interested reader can refer to a complete per- device, providing the desired electrical output, but these charges can spective on ion segregation for more details.36 Often, the changes also trigger changes in the material. Charge separation creates inter- are reversible in dark, but the light-induced changes can also be nal electric potential, which is often larger than the activation energy irreversible when crystal phase changes or chemical reactions or APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-7 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape are involved, leading to material decomposition. In situ character- TRPL along 6 h illumination in the humid air.44 The formation ization of an illuminated sample is an important tool to capture of PbI2 caused (i) parasitic absorption that reduced the perovskite the immediate effects of ion motion on the device and material PLQY below 520 nm and (ii) a gradual increase in TRPL lifetimes. degradation. The authors proposed that interfacial PbI2 is beneficial for defect passivation but bulk PbI2 provides detrimental non-radiative recom- 1. Single halide perovskites bination centers.44 Besides UV–vis illumination, perovskite is poorly stable under e-beam, posing an uncertainty for the characterizations One of the first light-soaking effects observed in situ was the using accelerated electrons. Yuan et al. combined PL and SEM imag- increase in PL emission intensity (also called photobrightening) with ing to demonstrate that the 10 keV e-beam quenched the PL signal longer laser irradiation.37–40 deQuilettes et al. found that the PL and reduced the film thickness by releasing volatile compounds.45 photobrightening of MAPbI3 films was correlated with an increase Interestingly, e-beam exposure prevented further film morphology in the time-resolved PL (TRPL) lifetime.37 The timescale of the degradation under illumination, indicating a complex relationship photobrightening in an isolated film was similar to the V oc rise between electrical currents and ion migration. of a PSC, suggesting a reduction in the bulk trap state density. The possibility of fabricating PSC on lightweight substrates ToF-secondary-ion-mass spectrometry (ToF-SIMS) depth profiling makes it appealing for space applications. Outside the protec- measured immediately after the laser excitation [Fig. 5(1)], provided tive Earth’s atmosphere, in operando conditions must include one of the first direct evidence of photo-induced iodide migration. harsh radiation. For that, Paternó et al. measured the J–V curve Even though this measurement was done on an isolated MAPbI3 evolution of MAPb(Ix Cl1−x )3 solar cells under visible light and film, its implications for the device’s stability were important. A neutron irradiation in a synchrotron.46 It was proposed that later work by Choi et al. spatially resolved the PL evolution in the neutron irradiation forms “benign” traps in the perovskite that polycrystalline film and the chemical composition by nanoscale- decrease the leakage current of the device, concluding that PSC resolved-ToF-SIMS.41 They found that higher oxygen content in are resilient to neutrons thus could potentially be used in space illuminated samples was correlated with greater PL lifetime and pro- applications. posed that light promotes O2− diffusion into the grains, passivating The substitution of all or part of the MA with less volatile lattice defects. organic cations, formamidinium (FA+ ) and Cs+ , provides an intrin- The existence of contradictory reports regarding the PL sically higher stable perovskite while retaining an appropriate Eg . 08 November 2023 15:51:28 response upon illumination (increase or decrease) motivated Systematic in situ XPS analysis comparing various combinations of Quitsch et al. to study the excitation-wavelength-dependence on the three cations confirmed the lower degradation rates for triple PL evolution, in air.38 Photobrightening occurred with long wave- cation perovskite upon illumination.47 However, perovskites with lengths, and a gradual trend to decreased emission toward PL mixed cation still suffer light-induced instabilities that have been quenching was observed at the 520 nm threshold (PbI2 band- less studied in the literature. Lu et al. studied compositional changes gap) [Fig. 5(2)]. The authors proposed that the process initiates by of α-FACsPbI3 films exposed to LED illumination with different photolysis of existing PbI2 impurities to Pb0 and I2 ,42 and I2 fur- excitation wavelengths (624, 520, 435, and 366 nm) in the XPS cham- ther decomposes MAPbI3 , quenching its PL.43 In vacuum and dry ber.48 While visible wavelengths resulted in a Cs+ migration toward air, only photobrightening at 405 nm was observed, suggesting that the bulk, UV light-induced degradation of the PbI6 4− into Pb0 and photo-decomposition requires the presence of moisture. Merdasa I− . A similar in situ XPS study on iodine-excess CsFAPbI3 revealed et al. tracked the film absorption, PL quantum yield (PLQY), and that 473 nm light-activated perovskite decomposition, catalyzed by FIG. 5. (1) ToF-SIMS depth profiling. (1a) Schematic of photon dose-dependent experiment indicating the regions exposed to 1.2 kJ cm−2 (red dotted circle) and 2.4 kJ cm−2 (green dotted circle). (1b) ToF-SIMS image of the iodide (I− ) intensity distribution summed through the film depth after local exposure to the photon doses shown in (1a). Adapted from deQuilettes et al., Nat. Commun. 7, 11683 (2016). Copyright 2016 Author(s), licensed under a Creative Commons Attribution 4.0 License. (2) Transient PL efficiency change for different excitation wavelengths. (2a) Time traces of the PL intensity of a MAPbI3 film sample for 60 s illumination at each excitation wavelength between 540 and 465 nm and measured in air. (2b) Extinction of a degraded MAPbI3 layer with increased PbI2 content in the spectral region around the PbI2 bandgap. Reprinted with permission from Quitsch et al., J. Phys. Chem. Lett. 9(8), 2062–2069 (2018). Copyright 2018 American Chemical Society. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-8 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape Br-richer phase. This finding had important implications for the device stability, as it implies a reduction of the maximum achievable V oc of the mixed-halide-based solar cells. Since then, much work was devoted to understand this phe- nomenon and all the variables that influence it, such as the A-site cation composition of the perovskite.51,52 Sutter-Fella et al. inves- tigated the reasons for the greater stability of the CsFA compared to MA perovskite for various I:Br ratios.53 Halide demixing (within 10–20 min) occur for both perovskites with high Br/I ratio, Fig. 7(1), with slight differences. While MA-based perovskite segregation happened in similar pathway to the report of Hoke, CsFA per- ovskite showed an intensity increase for both emission wavelengths, suggesting a reduction of the non-radiative recombination.53 In situ synchrotron XRD showed significant broadening of the MA- perovskite diffraction peaks and splitting due to formation of I-rich crystals [Figs. 7(1d) and 7(1e)]. In contrast, CsFA-perovskite showed just a small peak broadening [Fig. 7(1f)], i.e., the new I-rich domains are unable to relax the strain to attain a new lattice parameter, con- strained by the material around it. The hindered lattice relaxation would therefore be the key for higher stability in mixed cation/halide CsFAPb(Ix Br1−x )3 perovskite. A later work on a similar cation com- position [FA0.8 Cs0.2 Pb(I0.7 Br0.3 )3 ] demonstrated the suppression of halide segregation by incorporating potassium iodide (2 mol. %) in the lattice, perhaps due to a hindering on the ion migration due to the presence of K+ .54 While the scientific community reached a consensus that 08 November 2023 15:51:28 photo-excited carriers were the main driving force for halide segre- gation, the exact mechanism remained (and still is) elusive. Knight et al. demonstrated that halide segregation is not dependent on the accumulated number of photo-generated carriers, but rather on the number of charges that recombine through trap states.55 This was based on the observation of a much higher segregation rate (I-rich-peak vs mixed-cation-peak intensity ratio) when illuminat- ing a certain number of photons for long time (low intensity light, long exposure) than when providing the same photons but in a short time (high intensity light, short exposure) [Fig. 7(2)]. FIG. 6. (1a) Time-resolved in situ XPS spectra. Intensity map Pb 4f core-level spec- At high illumination intensities, the large number of generated tra with time. Dotted lines indicate centroid positions. (1b) XPS fit results for spectra marked with arrows in panel (2a) for blue-laser-exposed perovskite samples with carriers were enough to fill the traps, promoting mainly band-to- I/Pb = 3.8. Reprinted with permission from Donakowski et al., ACS Energy Lett. band recombination, while low photon flux results in substantial 6(2), 574–580 (2021). Copyright 2021 American Chemical Society. trap-mediated recombination that accelerates halide segregation. Therefore, to avoid segregation, the number of carriers recombin- ing via traps should be lower than 0.1%, providing clear guide- I3− [Fig. 6(1)]. Donakowski et al. suggested that photo- and ther- lines for the optimization of PSCs.55 In situ PL measurements on mal degradation could be prevented by adjusting the perovskite single-crystal CsPbBr2.1 I0.9 platelets confirmed that halide segrega- stoichiometry.49 tion routes depend on illumination power and atmosphere con- ditions, independently on single crystal or polycrystalline film.56 2. Mixed halide perovskites Low power density resulted in the simultaneous segregation of a As the perovskite research field matured, more complex halide pure-I-phase and a I-rich-phase, which later merged to an inter- perovskite stoichiometries provided greater stabilities than simple mediate I-rich phase after 25 min illumination. The authors pro- MAPbI3 , also enabling a broad Eg tuning, but had the issue of the posed that pure I-phase was formed at the crystal surface and halide segregation. This phenomenon was reported for the first time eventually reaches a thermodynamic equilibrium with the bulk in 2015 by Hoke et al.50 Several in situ PL works revealed the forma- forming intermediate segregated phase. When the measurements tion of the I-rich phase for various I/Br ratios on MAPb(Brx I1−x )3 were done in the presence of dry oxygen, the evolution of the perovskite, and some authors refer to it as “the Hoke effect.”51 While segregated phases became more complex, again emphasizing the the I-rich phase emission increases, the initial mixed-halide emission importance of surfaces in the segregation mechanism.56 Halide seg- peak diminishes, such segregation happens for all wavelengths above regation phenomena is not limited to polycrystalline films with the perovskite. It was hypothesized that smaller Eg I-rich domains abundant defects caused by the solution-based fabrication meth- act as carrier sinks for the injected charges from the wider-Eg ods. Mao et al. proved that halide segregation was an intrinsic APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-9 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 7. (1) Time evolution of steady-state PL spectra of FACsPb(I0.8 Br0.2 )3 (1a), FACsPb(I0.5 Br0.5 )3 (1b), and FACsPb(I0.4 Br0.6 )3 (1c) thin films taken at ∼50 mW/cm2 in 1 min increments. Synchrotron x-ray diffraction data and analysis of the [(1d) and (1e)] MAPb(I0.4 Br0.6 )3 and (1f) FACsPb(I0.4 Br0.6 )3 samples. (1d) [100] diffraction peaks before illumination, during, and after relaxation in the dark. (1e) and (1f) Normalized peak area under illumination over time. The inset in (1f) shows the Williamson–Hall plot of the XRD peak FWHM under illumination for 1 and 60 min. Adapted with permission from Sutter-Fella et al., Nano Lett. 18(6), 3473–3480 (2018). Copyright 2018 American Chemical Society. (2) Evolution of the PL intensity originating from the low-bandgap iodide-rich phase of MAPb(Br0.5 I0.5 )3 plotted against the number of incident photons per unit volume. Adapted from Knight et al., ACS Energy Lett. 4(1), 75–84 (2019). Copyright 2019 Author(s), licensed under a Creative Commons Attribution 4.0 License. (3a) Time-dependent fluorescence spectra of light-induced phase segregation recorded in the center of a single MAPb(Brx I1−x )3 microplatelet over 30 s, excited by a 400 nm pulsed laser (3.5 W cm−2 ). (3b) Widefield PL images of a MAPb(Brx I1−x )3 single crystal collected in the I-rich emission region (660–700 nm), excitation source of 400–450 nm, 5 mW cm−2 was used. Scale bar: 5 mm. Adapted with permission from Mao et al., Angew. Chem., Int. Ed. 58(9), 2893–2898 (2019). Copyright 2019 John Wiley & Sons. phenomena in perfect lattices by studying a single crystal of domains. To overcome this limitation, Halford et al. monitored the MAPb(Brx I1−x )3 perovskite with spatially and spectrally resolved photo-induced phase segregation using in situ grazing incidence in situ PL and TRPL.57 Upon illumination, two segregated PL peaks wide angle x-ray diffraction (GIWAXS) with ∼1 sun illumination appeared, similar to the observations on CsPbBr2.1 I0.9 single crys- for 1 h [Figs. 8(1a) and 8(1b)].58 Independently on the initial tals,56 but with the difference that the peak at ∼720 nm (I-richest) MAPb(I(1−x) Brx )3 halide ratios, the segregated phases had fixed com- did not blue shift, while the I-rich intermediate peak gradually red- positions, i.e., a majority of I-rich (x = 0.20) and (non-emissive) shifted [Figs. 7(3a) and 7(3b)]. Segregation was found to happen at Br-rich (x = 0.93) phases. The reversible remixing in dark [Fig. 8(1c)] the center of the crystal, proving that it does not happen necessarily was found to be substantially slower. Kim et al. employed a com- at the grain boundaries/crystal surface. bination of AFM, piezoresponse force microscopy (PFM), contact Despite the fruitful information uncovered by PL tech- potential difference (CPD), and XRD with in situ illumination and niques, they fail to identify the formation of non-emissive Br-rich biasing to reveal the microscopic structural changes produced in APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-10 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape FIG. 8. (1) In situ GIWAXS of MOHPs highlighting (1a) the experimental setup used to capture the dynamic structure of MOHPs under ∼1 sun illumination and (1b) MAPb(Br0.5 I0.5 )3 structure under 60 min of ∼1 sun illumination followed by 180 min in the dark. (1c) Dynamics of structural changes under illumination as shown via changes 08 November 2023 15:51:28 in diffracted intensity around the (200) perovskite peak for the MAPb(Br0.5 I0.5 )3 perovskite. Adapted with permission from Halford et al., ACS Appl. Mater. Interfaces 14(3), 4335–4343 (2022). Copyright 2022 American Chemical Society. (2a) 3D topography with (above) and without t (below) and corresponding line profiles (right) for topography under light (red) and dark (blue). The inset represents rate enhancement at the marked regions. (2b) Schematic illustration of the octahedral distortion induced by bias voltage and light illumination. Adapted from Kim et al., Nat. Commun. 10(1), 444 (2019). Copyright 2019 Author(s), licensed under a Creative Commons Attribution 4.0 License. (3a) HRTEM images acquired during 5 min (3b) characteristic diffraction patterns of the five abundant structures and (3c) corresponding abundance maps identified as CsPb(Brx I1−x )3 for x = 0.8, 0.45, 0.6, 1, as well as PbBr2 . Adapted with permission from Funk et al., J. Phys. Chem. Lett. 11(13), 4945–4950 (2020). Copyright 2020 American Chemical Society. the MAPb(Brx I1−x )3 film.59 Illumination caused an increase in the the LowDose synchrotron beamline for their photoemission spec- overall piezoresponse signal of the film, larger CPD at the grain troscopy (PES) measurements.61 PES analysis provided two main boundaries, and the formation of nanoscale ferroelastic domains in insights into the surface chemical changes of MAFAPb(I–Br) per- some grains of the film, seen also as increased surface corrugation ovskite illuminated with a 515 nm laser with varying energy. First, of ±3 nm [Fig. 8(2a)]. A reversible broadening of the FWHM was a gradual formation of Pb0 with illumination time, partly reversible associated with an increase in the strain disorder [Fig. 8(2b)]. This in dark, provided that there is I2 buried in the lattice. Second, an work found that the illumination induced mostly reversible changes, increase in Br− signal and reduction of I− and Pb2+ , suggesting while bias poling effects were not completely reversible. The irre- a phase separation into Br-rich regions at the surface and I-rich versible formation of amorphous material upon electrical biasing regions in the bulk. The authors proposed that Br-rich segregation was a probable causes of device degradation in PSCs.59 occur at the grain boundaries, contrasting with the works by Funk Funk et al. tracked the e-beam-induced phase segregation of et al. where single crystal also showed Br-rich segregation in the crys- CsPb-halide perovskite nanocrystals grown on a TEM grid by using tal center.60 Probably the different perovskite cations (MAFA vs Cs) high resolution (HR)-TEM.60 Multivariate analysis methods enabled and stress sources (visible laser vs e-beam) used in the works of Cap- the (spatially resolved) deconvolution of diffraction patterns for Br:I pel and Funk, respectively, originate diverse mechanisms for halide ratios 80:20, 45:55, 60:40, 100:0 and PbBr2 [see Fig. 8(3)]. Substan- segregation. Future works would benefit of exploring various per- tial changes appeared the first 5 min of irradiation, i.e., large parts ovskite compositions to ultimately obtain a universal model for the of the crystal became amorphous, and the original 80:20 uniform phase segregation phenomena. composition changed to pure Br-perovskite in the crystal center Among the in situ reports found in the literature, we observed and predominantly 60/40 (I-richer) around the edges. Note that that a vast majority of experiments use steady-state PL, TRPL, Br-rich phases could not be detected before by solely using PL meth- and photo-induced absorption to track light-induced changes ods. The authors hypothesized that the asymmetry in ionic bonding in the optoelectronic properties, taking advantage of the laser strength of I and Br with Pb caused lattice strain that relaxed under excitation of the instrument to perform light soaking.50,52,62–67 irradiation (e-beam or light) creating separated phases. Cappel Other reports complemented the knowledge on the material crys- et al. minimized the x-ray damage on the perovskite by employing tallinity and composition with in situ XRD techniques and x-ray APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-11 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape photoelectron spectroscopy (XPS), respectively, by adding a light on stress-level applied and the measurement environment. An inter- source through the windows of the instrument chambers.50,68–72 mediate PbI2 -like structure acts as the base for PbI2 formation, Other less readily available in situ techniques, such as ToF-SIMS, similarly to how this intermediate seems to act as base for MAI inter- TEM, synchrotron beamlines, or custom-made setups, provided calation. Therefore, thermal degradation is often called the reverse valuable insights into complete the overall knowledge on perovskite of perovskite formation. Breakage of the Pb–I–Pb bonds in the [001] light-induced degradation. direction leads to the formation of an intermediate with channels, allowing for increased migration of species throughout the material. IV. THERMAL STRESS This allows for metallic lead (Pb0 ), iodide, I− , CH3 NH3 + , HI, and CH3 NH2 to quickly migrate and either diffuse or implant inside Temperature is an unavoidable stressor in photovoltaic appli- the perovskite grains or interfaces. Intermediate structures or even cations since sunlight radiation includes infrared heating energy. amorphization of the material arises to start recrystallization that According to international protocols,6 heating between 60 and 85 ○ C stops at the more stable PbI2 . The creation of PbI2 during heating of should be considered normal during operation, yet margins for MAPbI3 seems to arise from crystal instability caused by the organic higher temperature endurance are important to withstand applica- cation. Under isolated, low thermal stress (<130 ○ C), the reaction tions in hot climates or incorporated in technology that might get is a layer-by-layer transformation starting with dangling bonds on warm under operation. the surface of perovskite crystallites. Layer-by-layer decomposition There is a consensus in the literature that PbI2 is the product of mechanisms can be slowed down by passivation of dangling bonds thermal degradation of MAPbI3 , yet the degradation path depends or increased integrity of the crystal at surfaces and grain boundaries. 08 November 2023 15:51:28 FIG. 9. (1a)–(1c) Degradation process of an individual MAPbI3 grain. The purple dashed lines outline the shrinking perovskite grain as it degrades to PbI2 . Insets: FFT phase diagrams of the corresponding HRTEM images. Scale bar, 2 nm. (1d) Transition from MAPbI3 with a tetragonal configuration to PbI2 with a trigonal configuration. Reprinted with permission from Fan et al., Joule 1(3), 548–562 (2017). Copyright 2017 Elsevier. (2a) HRTEM images and FFTs obtained from the in situ HRTEM, demonstrating the detailed intermediate process. Critical alterations, emergence of black dots, and splitting of diffraction spots (white arrows in the red rectangles) are observed from ≈130 ○ C (2b) Simulated cross-sectional thin plate of a perfect trigonal PbI2 crystalline along [412]. Reprinted with permission from Kim et al., Adv. Funct. Mater. 28(42), 1804039 (2018). Copyright 2018 John Wiley & Sons. (3) HRTEM of (A) fresh caffeine-containing PVSK, (C) fresh pure PVSK, (E) aged (5 min 30 s) caffeine-containing PVSK, and (G) aged (5 min 30 s) pure PVSK. Corresponding fast Fourier transforms (FFTs) of (B) fresh caffeine containing PVSK, (D) fresh pure PVSK, (F) aged (5 min 30 s) caffeine containing PVSK, and (H) aged (5 min 30 s) pure PVSK. Adapted with permission from Wang et al., Joule 3(6), 1464–1477 (2019). Copyright 2019 Elsevier. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-12 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape At higher temperature stress (130–150 ○ C), the reaction happens fast they indicate, on cross-sectional samples, particles may migrate on and seem to pass by a more amorphous soup containing the per- the surface of the sample along the cross-section, i.e., the exposed ovskite components. Degradation of MAPbI3 seems to occur at all surface will be an additional migration path that do not exist in an elevated temperatures, although the stress seems partly reversible actual device and FIB milling may induce defects seen in the experi- until final formation of PbI2 . For mixed cation perovskite composi- ment. Still, using a micro-heater in situ stage in TEM, they outlined tions, thermal degradation becomes considerably more complicated. the requisites for obtaining energy dispersive x-ray (EDX) mapping Due to the atomic radius differences, there seems to be an increase in in situ. Heating from 50 to 150 ○ C resulted in iodine and lead migra- integrity of the crystal lattice. In situ electron microscopy has allowed tion as a part of the degradation process, but the rate and quantity of for video-recordings of thermally induced degradation in real-time, such migration occurred differently depending on air exposure dur- and videos can be found in the supporting information of several ing fabrication, and the device fabricated in 50% RH air was more articles.73,74 Because of the strong dependence of thermal stability susceptive. For the other samples, the degradation seen in TEM at on the crystalline composition, we will first discuss the stability of 200 ○ C was comparable with degradation seen after storage for a MAPbI3 followed by mixed cation and mixed halide compositions replica device in air for 2 months. Increased susceptibility to heat as seen by in situ characterization methods. stress depending on air exposure has also been related to oxygen, as Fan et al. used in situ TEM to monitor the reciprocal selected- shown by Yang et al.77 area electron diffraction (SAED) signal from the crystalline structure Moving on to mixed composition perovskite, Seo et al. reported of a MAPbI3 microplate, while it was subjected to 85 ○ C in a heated different thermal-degradation phenomena for devices cross sec- environmental gas cell.75 The initial preferential orientation of the tions, including MAPbI3 and Cs0.175 FA0.750 MA0.075 Pb(I0.880 Br0.120 )3 , crystal [001] still remained after applying a thermal stress (85 ○ C), within device architecture ITO/mp-NiOx /perovskite/PCBM/mp- but after 60 s, the hexagonal pattern corresponding to PbI2 appeared ZnO/Ag.78 It was the first study of in situ TEM thermal degradation [Figs. 9(1a)–9(1d)]. After only 100 s, 75% of the initial hexagonal of triple cation PSC. The perovskite film samples were held for MAPbI3 material was estimated to be converted to trigonal PbI2 phase [Figs. 9(1c) and 9(1d)]. They found that PbI2 formed simi- lar to the reverse process of formation of MAPbI3 via intercalation of MAI into PbI2 .76 The conversion showed every indication of being directly sequential and persistence of initial peaks through- 08 November 2023 15:51:28 out the transition proved that the periodicity of the [400] direction in MAPbI3 coincides with the {210} facets of the PbI2 structure after transitioning. It is suggested that heating stimulates breakage of the [001] Pb–I–Pb bonds, promoting the release of CH3 NH3 + and I− ions in the form of CH3 NH2 and HI gas. PbI2 forms immediately but is held in an intermediate lattice structure first (keeping the lat- tice spacing of the MAPbI3 [400]) before transforming into regular trigonal PbI2 (at 420 s). The degradation pathway was the same inde- pendently of using dry air (2% RH, 700 Torr) or vacuum (0% RH, 1 Torr), and density functional theory (DFT) calculations from the in situ SAED data confirmed the reaction and that the degradation was driven from the surface of the crystallite in a layer-by-layer man- ner.75 In contrast to the results presented by Kim et al., Fan et al. used high resolution TEM (HRTEM) to report an intermediate step in the PbI2 formation via an amorphized MAPbI3 layer.74,75 Instead of using a controlled heating stage,75 the sample was heated with the electron beam from the TEM. The sample was estimated to be heated from 25 to 170 ○ C, Fig. 9(2a), where a gradual change of MAPbI3 diffraction peaks [by Fast Fourier Transform (FFT)] into amorphous rings started at 70 ○ C (2 min 15 s). This was followed by the for- mation of crystalline black dots from 130 ○ C (3 min 5 s) onwards, seen nucleating on the sample with amorphized surroundings on the TEM image. The intermediate phase shows oriented channels that could enhance ion transport and facilitate the formation of the PbI2 [Fig. 9(2b)]. Together these two works indicate that, with high irradiation, amorphization may occur due to the faster breaking of Pb–I–Pb bonds, fast segregation of the products led to more con- densed crystallites of PbI2 growing from the amorphous phase. The FIG. 10. (1) STEM images of (1a) MA-, and (1b) Cs/FA/MA- based PSC sample growth of the PbI2 grains still showed preference for growth toward and EDS elemental mapping images of lead (green) and iodine (yellow) for (1c) the surface.74 As a last example of in situ TEM on MAPbI3 , focused MA-, and (1d) Cs/FA/MA- based PSC sample, after heating from RT to (85, 130, ion beam (FIB) cross sections (CS) of full solar cell devices with four 150, and 170) ○ C. Reprinted with permission from Seo et al., Nano Energy 77, different fabrication methods were compared by Divitini et al.73 As 105164 (2020). Copyright 2020 Elsevier. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-13 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 30 min at 85, 130, 150, and 170 ○ C, and in-between each step, sam- by coupling a quadrupole mass spectrometer (MS) to the exhaust of ples were cooled down and electron dispersive x-ray spectroscopy the heating chamber when monitoring the level of degradation of (EDS) maps were acquired for 10 min. The EDS elemental distri- MAPbI3 with thermal gravimetric and differential thermal analysis bution (Fig. 10) revealed that the MA-based device degraded partly (TG-DTA) during heating. They could observe MA loss at 294 ○ C at over 85 ○ C, with gradual growth of defect phases in the MAPbI3 composed of two sections, first decomposition of MAPbI3 and then layer [Fig. 10(1a), 130–170 ○ C]. The Cs/FA/MA-based device, in decomposition of MAI, but during both steps, the only products contrast, resisted degradation up to 130–150 ○ C, afterward rapid detected were CH3 I and NH3 . In a later study, they demonstrated formation of large metallic lead and PbI2 particles. There was also that, at temperatures relevant for PSC operation (<100 ○ C), FA- some gradual degradation of the Ag electrode for the MA device, based perovskite released degradation products to much lesser scale not observed for the Cs/FA/MA-based device. As redistribution of than MA-based. For FA-based samples, they detected signals from iodide and Pb happened at much lower temperatures in MAPbI3 sym-triazine (C3 N3 H3 ), formamidine [CH(NH2 )2 + ], and hydrogen [Fig. 10(1c)] than in Cs/MA/FA [Fig. 10(1d)], the authors con- cyanide (HCN) in very different amounts depending on the tem- cluded that activation energy of ion migration was lower in MAPbI3 . perature applied to the sample. Such degradation products originate Mixed cation perovskite, however, degraded considerably faster in irreversible reactions. Noticeably, the heating rate did not have with more severe morphology deformation [Fig. 10(1b), 170 ○ C]. any apparent effect on the chemical reactions that occurred. Kim It was hypothesized that the thermally induced Br− vacancies ini- et al. used in situ surface analysis techniques to also confirm that tiated the fast creation of metallic lead.78 A similar hypothesis was thermal degradation of MAPbI3 happens through an intermediate drawn by Long et al. by in situ AFM for mixed cation and halide forming end products PbI2 , CH3 I and NH3 .88 This supported the perovskite.79 Aguiar et al. studied FAPbI3 decomposition mecha- results obtained by mass spectrometry by Juarez-Perez et al.86,89 It nisms with in situ STEM, SAED, and in situ XRD. They reported was shown that encapsulated PSC devices were not protected against nucleation and growth of precipitates containing high levels of lead thermal degradation during operational stability tests at 85 ○ C. (PbI2 , Pb0 ) above a 175 ○ C temperature threshold.80 In addition, In situ surface analysis with 2D GIWAXS, HR-XPS, and near edge with separate in situ XRD and in situ XPS measurements, Long x-ray absorption fine structure (NEXAFS) under vacuum to perform et al. concluded that the FA+ and Br− ions lose their interaction heat-stress tests allowed screening the top 125 nm surface a MAPbI3 with the perovskite crystal during thermal annealing, resulting in film as it transformed from tetragonal to cubic phase upon increas- a phase richer in I− than what was intended during fabrication. A ing temperature from 25 to 80 ○ C in 20 min. The process was fol- 08 November 2023 15:51:28 lower angle shift of the (100) perovskite peak of Cs0.17 FA0.83 PbI2 Br lowed by the creation of the intermediate structure mentioned above along with increased temperatures. The authors hypothesized that at 100 ○ C for 20 min. Note that this time again the PbI2 planes were the shift was due to Br-loss, since the Br atom occupies less space in kept with longer inter-planar distance than normal PbI2 . Finally, the the lattice than iodide, it led to an expansion.79 Thermal expansion crystalline PbI2 degradation-product appears as the other volatile of the lattice could originate a similar XRD observation, accord- degradation products are removed. Extended exposure to lower tem- ing to other works.81,82 Additionally, confirming their hypothesis, peratures (80 ○ C) for over 60 min produced similar changes. Finally, the interaction of Br+ and FA+ ion extracted by XPS changed at they varied the angle of incidence from 0.1○ (5 nm depth) to 0.15○ much lower temperatures (100 ○ C) than iodine. Borchert et al. the (55 nm depth) to 0.2○ (125 nm depth) to confirm that the surface Cl-containing MAPbI3 film improved its crystallinity until 200 ○ C, of the film had been more degraded. The authors argue that, in followed by a rapid decrease and the appearance of PbI2 and encapsulated devices, the degradation would happen in a different metallic lead, seen with in situ XRD.83 Compared to other authors pathway since the volatile degradation products may accumulate at who performed tests in nitrogen atmosphere,84 the faster decompo- the interface or migrate back into other layers of the PSC.88 Subse- sition in vacuum was related to the removal of gaseous products. quently, in situ HR-XPS on the top 10 nm of a MAPbI3 film heated For the pure MAPbI3 film, the results coincided with future find- over 100 ○ C confirmed MA release from the film. These changes ings, i.e., cubic phase formation at ∼50 ○ C, a later increase in crys- appear at much lower temperatures than those indicated by Juarez- tallinity followed by various unidentified peak changes starting at Perez et al. (∼200 ○ C),86 suggesting that MAPbI3 is susceptible to 150 ○ C.74,85,86 It is likely that the ionic radius of the Cl− also played heat stress on the outermost surface much earlier than the bulk. It a role here. Indeed, the exact perovskite stoichiometry greatly influ- makes sense that decomposition products are detected later as the ences the ease in which it degrades under thermal stress. A good surface/volume ratio in a thin film is very small and CH3 I and NH3 example is the different activation energies, ∼0.66 and ∼0.76 eV evaporate at trace amounts. It is also proposed that the high vacuum reported for the thermal degradation of Cs0.17 FA0.83 Pb(I0.83 Br0.17 )3 conditions accelerated the thermal degradation. Finally, in situ near and Cs0.05 (MA0.17 FA0.83 )0.95 Pb(I0.83 Br0.17 )3 films, respectively. These edge x-ray absorption fine structure (NEXAFS) studies in a similar rough values were found by Tan et al., thanks to in situ XRD with setup gave insight to the interaction of molecular MA+ ions with the temperature controlled stage.87 It is a clear indication that slightly intermediate structure before PbI2 formation. The linear polariza- lower Cs+ concentration has important stabilizing effects at elevated tion of the synchrotron x-rays was used to probe the orientation of temperatures (up to 150 ○ C), despite the presence of volatile MA the C–N bonds during increasing temperature at UHV conditions. ions. When subjecting the sample to similar conditions as when changes Juarez-Perez et al. identified that the majority of thermally appeared by in situ 2D GIWAXD (100 ○ C for 20 min), they could decomposed gas-products of MAPbI3 crystals were CH3 I and NH3 , observe how CH3 NH3 + ions aligned themselves in an orthogonal contrary to some previous reports that suggested that decomposi- fashion with the substrate surface. Over time, during heating, there tion products were CH3 NH2 and HI.86 Their experiments were done was also a drop in intensity of the peak corresponding to the C–N APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-14 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape bond. Results indicate that CH3 NH3 + are more aligned with hydro- (ii), which is the precursor of PbI2 formation [Fig. 9(3)]. Further gens pointing toward the PbI2 sheets of the intermediate structure exposure to the e-beam (up to ∼135 ○ C) caused the decay of (110) before the decomposition into CH3 I and NH3 .88 To present fur- diffraction peak for both samples, but while new diffraction peaks ther support of threshold temperatures for degradation, Wang et al. were observed in the control sample, the caffeine-containing sample monitored the temperature-dependent degradation of MAPbI3 by did not undergo such new peak formation. It was hypothesized that in situ spectroscopic ellipsometry (SE) to track the PbI2 content, film the caffeine acted as a molecular lock via strong interaction with the thickness and dielectric constant over time.90 Full degradation was Pb2 + ion, preventing the formation of amorphous phases, thereby defined as when a dielectric constant no longer changed, happened preventing the degradation at elevated temperatures. Another strat- at different timeframes, inversely proportional to the temperature egy was reported by Fan et al., demonstrating improved thermal heating, i.e., 3.8, 14.5, 39.7, and 137.7 min at 220, 200, 180, and stability with hexagonal boron nitride (hBN) thin flakes in hBN- 160 ○ C, respectively, following an Arrhenius equation. Unsurpris- perovskite-hBN heterostructures.75 The B+ and N− ions in the flakes ingly, the MAPbI3 thickness decreased and PbI2 content linearly passivate the dangling bonds of the MAPbI3 surface, reducing its increased over time. susceptibility to rearrangement during the thermal stress. This does Included above are several good examples on how tuning the not completely remove thermal instability but reduces the reactivity stoichiometry can make more robust MOHP against elevated oper- of the surface and prolongs the time for layer by layer degradation ating temperatures. Other strategies to increase the resistance are seven-fold. the incorporation of additives in the perovskite film. In the work of A last note is that in operando stressors can induce lattice Maniyarasu et al., 1-octyl-3-methylimidazolium chloride ionic liq- expansion of crystalline materials. Changes in lattice spacing mod- uid (IL) hindered FA0.9 Cs0.1 PbI3 perovskite thermal degradation in ify the material dielectric constant that, in turn, can be detected by UHV.28 The findings were obtained via in situ near ambient pressure time-resolved and steady-state optical reflectivity (Fig. 11). Li et al. (NAP)-XPS at room temperature (RT), 100 ○ C, and 150 ○ C, com- employed the technique to study lattice expansion on MAPbI3 single paring a reference and IL-modified sample. While the IL-modified crystal under continuous Xe-lamp illumination and various temper- perovskite showed minimal changes at 100 and 150 ○ C, increased atures.81 It was observed that illumination by itself did not induce signals of Pb0 and PbI2 originated in the reference sample improved any change in the lattice. Instead, it was found that the heating crystallinity and reduced the presence of unwanted Pb0 . The authors caused by the illumination source was the main factor causing ther- hypothesized that the IL helped improve the perovskite crystallinity mal lattice expansion, similar to heating the sample in the dark. The 08 November 2023 15:51:28 and prevented FA+ ions from diffusing out of the perovskite crys- contrast with the above studies that found substantial ionic redistri- tal lattice. Wang et al. increased the thermal stability of MAPbI3 by bution upon illumination relies perhaps in the low irradiation power incorporating the authors own driving force molecule, i.e., caffeine, used in the reflectivity studies. in the perovskite film.91 Their observations of in situ HRTEM crys- To conclude this section of in situ thermal degradation studies, tal evolution agreed with previous findings by Kim et al.,74 MAPbI3 we extract a general consensus in which the main degradation prod- decomposition happen in the sequence of (i) amorphized material, uct is PbI2 , with the release of volatile organic molecules and halide (ii) intermediate crystalline phase, and (iii) PbI2 . Indeed, improved if the device is not encapsulated.92 For MAPbI3 , the decomposition photovoltaic stability at 85 ○ C under nitrogen flow was obtained. happens through a crystal change that transitions from amorphous Caffeine molecule hindered the formation of the intermediate phase to a precursor of the PbI2 phase. Relative to the protection strategies FIG. 11. (1a) Schematic diagram of time-resolved optical reflectivity measurement based on ultrafast pump–probe method. The inset shows the sample photographs of single-crystalline MAPbI3 perovskite and temperature control device with control accuracy from 4.5 to 320 K. (1b) 2D mapping diagram of reflectivity, where the black arrows represent the thermal expansions of lattices. Adapted from Li et al., J. Appl. Phys. 132(1), 013102 (2022) with the permission of AIP Publishing. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-15 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape adapted to increase resilience to temperature stress, we can mention voids, migration of iodide into the HTL, and the volatilization of the precise tuning of perovskite stoichiometry, the incorporation of iodide and organic molecules under 6 V forward bias by STEM additives in the film, or even the addition of surface modifiers to and HAADF.94 They observed strong preference for degradation create more robust interfaces. at the MAPbI3 /HTL interface, believed to be caused by an energy- We would like to point out that several of the reported in situ dissipation process depending on the polarity of the layers at their measurements employed illumination or e-beam as the heating intersection. In a TEM stage allowing both heating and electrical energy that indeed cause important temperature increase. High biasing, Kim et al. imaged the amorphization of a 300 nm lamella light-induced temperature is an overlooked stressor that may influ- (prepared by FIB) of a complete PSC, suspended between two Au- ence the degradation mechanisms in experiments performed under electrodes, the first 5 min of 1 V forward biasing [Figs. 12(2a) and illumination, especially Xe-lamp containing IR fraction of the spec- 12(2b)].95 Under +1 V electrical bias, they could observe by HRTEM trum, or high-power lasers. Often, sample heating is not consid- how perovskite grains lost crystallinity over time, as if the crystalline ered when providing a hypothesis of the light-induced degradation layer was dissolved [Fig. 12(2c)]. PbI2 phase present did not lose its mechanisms [see Sec. III (illumination stress)]. crystallinity. No morphological changes were observed, and changes Measurements on complete solar cell devices are desirable could only be seen when reaching lattice spacing resolution. To dis- since they incorporate the effects of the adjacent selective contacts; card beam-damage, the results seen in HRTEM were also verified however, some techniques are rather limited in this regard. For with SAED before and after 5 min at 1 V on a spot of the lamella that example, the creative authors doing FIB cross sections of full PSC was not measured by TEM during the biasing [Fig. 12(2d)]. With for TEM imaging themselves report FIB cutting as a possible factor SAED they could see clear formation of amorphous rings after the of degradation. biasing [Fig. 12(2e)]. Interestingly, the device recovered fully by rest- ing in the dark, and also complete recovery was achieved by heating to 50 ○ C in merely 3 h. The authors proposed a mechanism of amor- V. BIAS VOLTAGE-INDUCED DEGRADATION phization via halide ion migration and defect generation that leads In commercial solar panels, photovoltaic cells are connected in to crystal collapse as the Pb-plane decomposes. Despite sophisticated series to achieve greater output voltages. Generally, panels operate methods, Kim et al. could provide simple and clear conclusions, close to the maximum power point (MPP) if all individual cells have thanks to the separation of bias, e-beam, and temperature stress.95 similar efficiencies and illumination all will work at similar maxi- Their results are revolutionizing regarding the increased under- 08 November 2023 15:51:28 mum power point voltages (V MPP ). V MPP is a voltage usually close standing of perovskite degradation under bias conditions but also to V oc , which when applied produces the highest product between the very promising healing of bias stress damage by exposure to mild voltage and current (power), i.e., where the maximum amount of elevated temperatures. photogenerated energy is extracted through the contacts. Electri- Functionalized atomic force microscopy (AFM) has been used cal charges that are not extracted and accumulate in the device can in several studies to spatially resolve the contact potential differ- trigger instabilities, e.g., in the presence of O2 , free electrons can cre- ence (CPD), Kelvin probe force microscopy (KPFM), piezoresponse ate reactive superoxide species (O2 − ) that degrade the perovskite.31 (PFM), and conductivity (c-AFM), simultaneously as the surface Also, partial or total shading of panel-units may force shaded cells topography. These techniques have been used in several works, to operate in reverse bias to accommodate passing electrical current. with both lateral contact samples and cross-sections devices, to Reverse biases induce high electric fields that can initiate unwanted apply electrical bias and prove the existence of ionic migration and ion-drift with consequences on the perovskite stability. Eventually, accumulation at interfaces of the active layer.59,96–98 The voltage dis- reverse bias leads to breakdown voltages that would destroy the tribution is found to be inhomogeneous and variable across grains device. under illumination.59,99,100 With in situ AFM methods and the varia- We will also review findings from in situ stability studies apply- tion upon applying/removing bias, one can obtain insights about the ing bias voltage stress to complete PSC devices or thin films with ion distribution over time. Using KPFM, Ahmadi et al. investigated symmetric electrodes. First, we will focus on the studies that describe the interfaces of a single crystal with lateral Au electrodes under bias the forward bias or symmetric electrode bias, and second, the reverse to probe the dynamic charge changes.96 For short times, the authors bias effect in complete PSC, i.e., diode-like structure. saw indication of slow changes in charge-distribution produced by Li et al. presented an important in situ PL study to obtain higher bias (4 V) conditions. Lower biases (1 V) were not sufficient quantitative measurements on the speed of ionic motion. A bias to activate the motion of slow ionic systems or other charged species. (∼2 ⋅ 104 V m−1 ) was applied between two parallel electrodes sep- The dynamics in motion were clearly bias-dependent. Given the fact arated 150 μm, and this produced a parallel front of darkened PL that these phenomena happened at the same time, they proposed intensity from the positive to the negative electrode, stretching out that interfacial charge injection and ion migration may be intrinsi- over time [see Fig. 12(1a)].93 By tracking simultaneous current loss cally linked. The work by Kim et al. discussed in Sec. III [Fig. 8(2b)] along the bias stress [Fig. 12(1b)], the authors determined the ion found that illumination induced mostly reversible changes in a migration speed as 10 μm/s. The defects induced by the front drasti- MAPb(Brx I1−x )3 film, while bias poling effects were not completely cally limited radiative recombination in the film between the lateral reversible.59 From in situ XRD, peak broadening during biasing was contacts. irreversible, contrary to reversible broadening from illumination. The development of advanced TEM stages has allowed for Additionally, they observed that +2 V induced contact potential dif- visualizing the MAPbI3 degradation under bias. Jeangros et al. ference (CPD) variations within grains while no variations were seen concluded that the main culprits of bias-induced degradation of under −2 V, suggesting a better charge separation at +2 V than MAPbI3 solar cells were the formation of PbI2 nanoparticles and at −2 V. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-16 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 12. (1) Observation of a dark front in a light-soaked perovskite. (1a) Time-dependent PL images of a perovskite film CH3 NH3 PbI3−x Clx under an external electric field (∼2 ⋅ 104 V m−1 ). The “+ ” and “−” signs indicate the polarity of the electrodes. The channel length is ∼150 μm. The scale bar represents 100 μm. Blue arrows, →z, indicate area of PL lost since bias application (1b) Electrical current 1/I2 monitored as a function of time during the measurement of experiment (1a). Adapted from Li et al., Nat. Commun. 9(1), 5113 (2018). Copyright 2018 Author(s), licensed under a Creative Commons Attribution 4.0 License. (2) In situ observation of changes in perovskite materials induced by electrical biases. (2a) Schematic of in situ TEM sample configuration under electrical bias. (2b) SEM image of the nano-solar cell lamella used for in situ observation prepared by FIB. The scale bar is 2 μm. (2c) HRTEM image series of perovskite materials under 1 V electrical bias. (2d) TEM image of nano solar cell indicating the locations of the SAED area. (2e) SAED patterns of perovskite materials at the highlighted region after 5 min of forward bias (1 V) without beam radiation. Reprinted with permission from Kim et al., ACS Energy Lett. 6(10), 3530–3537 (2021). Copyright 2021 American Chemical Society. (3a) Layout of the electrodes used for the investigated PSCs, designating an active area of 1 cm2 . (3b)–(3f) Temperature maps measured for a PSC with Li-doping of m-TiO2 at the rear side during operation at increasing reverse biases, from 5 to 25 V. Reprinted from Najafi et al., ACS Appl. Energy Mater. 5(2), 1378–1384 (2022). Copyright 2022 Author(s), licensed under a Creative Commons Attribution 4.0 License. Liu et al. employed time-resolved (tr)-ToF-SIMS and tr-KPFM dark or illumination conditions. The electric field induced further to track the in operando ion migration and evolution of local charge decomposition of CH3 NH3 + into various sub-products, especially density in MAPbI3 between parallel Au electrodes [Fig. 13(1)].101 in dark conditions and at the interface with the electrodes. Ni Direct evidence of CH3 NH3 + and I− migration toward opposite et al. employed drive-level capacitance profiling (DLCP) and ther- directions was found when biasing the electrodes, independently of mal admittance spectroscopy (TAS) to determine the change on trap APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-17 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape state density and position induced by reverse electrical poling of a e.g., carbon-based electrodes can increase resilience against reverse PSC, a common scenario in partially shaded PV modules.102 A pro- bias. gressive density increase of two trap-bands upon continuous biasing The contacts are of high importance when performing bias- at −1 V [Fig. 13(2)] was attributed to iodide interstitials near the stress stability tests. Characterization using film-contact methods to C60 /perovskite interface. apply bias are sensitive, e.g., Ahmadi et al. reported that the Au- As stated, reverse biases are relevant to series-connected cells contact interface with the perovskite is non-ohmic, thus creating in a module from the well-known effect of partial shading. There- interfacial polarization and recombination that could play a role in fore, several works attempted to study the influence of reverse bias degradation of the film.96 Metallic contacts can create filament chan- on device stability. Najafi et al. monitored a small area with infrared nels that cause shunts in a device, or metal-halide reactions, such thermal imaging together with the current density response of a as AgI creation, can occur. Carbon-based electrodes are an alterna- silver-electrode PSC [Fig. 12(2a)].103 The device architecture was tive to mitigate some bias-induced degradation by substituting the found to be vital for resilience against high reverse bias, something metal. previously inferred by other authors seeing that many irreversible Overall, most works mention unwanted charged species (e.g., reactions during biasing occur at the device interfaces.104 As reverse ions) motion and crystal amorphization as a main cause for per- bias was applied gradually from 2.5 to 30 V the temperature of the ovskite degradation under forward bias application. Reversibility of device increased from 18 to over 160 ○ C, indicating the formation some changes once the electric field is removed is observed, but high of shunting pathways that degrade the device [Figs. 12(2b)–12(2f)]. bias can induce irreversible decomposition reactions and even inter- In the study, the authors compared three variations of the FTO/m- diffusion of the metal electrode across the device layers. Despite TiO2 /perovskite/spiro-OMeTAD/Au architecture, with and without the fact that reverse bias damage in commercial solar panels can compact TiO2 (c-TiO2 ) and with Li-doped mesoporous TiO2 (Li- be solved through electrical design adding bypass diodes, it is rel- m-TiO2 ). The absence of c-TiO2 layer accelerated the temperature evant for other potential applications such as memristors where damage induced by reverse bias. Li-doping, however, was found to negative biases can be beneficial. We note that upon applying high be an effective method that improves PSC performance while leaving biases, changes can occur fast (<2 s) and techniques are limited by stability against reverse bias unaltered. The authors proposed that the time required to complete a full scan, such as KPFM. To see the origins of reverse bias degradation on the Au electrodes included such fast changes, a higher time-resolution needs to be developed, local shunts from metal ion migration and local heating from or results obtained by these methods are analogous to “before and 08 November 2023 15:51:28 arc faults. It is therefore suggested that using non-metal contacts after images” done in situ under controlled conditions. FIG. 13. (1a) Schematic of ToF-SIMS measurement on lateral gold electrodes on a CH3 NH3 PbI3 film. (1b) Schematic shows possible ions existing in CH3 NH3 PbI3 , black box shows intrinsic ions including CH3 NH3 + and I− , and green box shows CH3 NH3 + decomposition products. Adapted with permission from Liu et al., ACS Nano 15(5), 9017–9026 (2021). Copyright 2021 American Chemical Society. (2a) Schematic diagrams of the device structure of a MAPbI3 thin single-crystal solar cell under reverse bias (left) and the movement of charged ions under electric field E (right). (2b) Trap density of states (tDOS) spectra before and after applying a reverse bias of −1 V with different durations. Spatial trap density distribution for trap-band (2c) I and (2d) II in the MAPbI3 thin single crystal solar cell after applying the reverse bias measured by DLCP. The inset in (2c) displays the zoomed-in trap density change near the C60 /perovskite interface. Adapted with permission from Ni et al., Nat. Energy 7(1), 65–73 (2022). Copyright 2022 Nature Publishing Group. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-18 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape VI. CYCLE-APPLIED STRESS a reproducible full recovery after dark of the two emission peaks monitored (1.68 and 1.94 eV) [Fig. 14(1b)].50 This light-induced, While previous sections II–V reviewed the individual stress fac- reversible phenomenon was a seminal work for cycling studies. tors of humidity, oxygen, illumination, temperature, and bias, we Complete reversibility of light-induced halide segregation in prop- now begin exploring the combination of such conditions to better erly protected environments has been argued in further works. In understand the long-term degradation of PSC in actual operation. the study by Knight et al., in situ PL spectra of MAPb(I0.5 Br0.5 )3 Testing reversibility and changing conditions (e.g., day/night) dur- films were collected along eight cycles of 15 s illumination followed ing stability assessment of solar cells is relevant to mimic outdoor by 30 min of dark rest, in four different environments: air, vacuum, operation. Standardized ISOS protocols include various cycling tests nitrogen, and encapsulated (PMMA coating done in vacuum before (T, LT, LC) to account for the cycles, e.g., day and night, rain and any environmental exposure).55 Under inert nitrogen or encapsu- sun, hot and cold, etc., of long term operation in PV installations. lation [Figs. 14(2c) and 14(2d)], similar PL behavior was obtained Changes in conditions may lead to partial recovery resulting in low under illumination, each time for eight cycles. In vacuum and air overall degradation or lead to destabilization of steady-state condi- [Figs. 14(2a) and 14(2b)], there were changes over time in opposite tions and acceleration of performance decay. Most commonly, some directions for the conditions. In air, the signal from iodide-rich per- irreversible part of PSC degradation accumulates over time during ovskite reached higher intensities than in other conditions, and the repeated stressing and relaxation. photobrightening was faster each cycle. In vacuum, the iodide-rich Some works already mentioned in this review complemented photobrightening happened at slower rates with additional cycles their experiments with cycling. In the original report of the “Hoke and the mixed-phase PL signal decreased exponentially over time effect” in 2014, halide segregation was observed as a shift of PL emis- independently on the illumination cycles. This suggested that, in sion to lower energies [Fig. 14(1a)]. A MAPb(Br0.6 I0.4 )3 film during both vacuum and air, irreversible processes take place under illumi- four cycles of 2 min of illumination and 5 min of dark-rest showed nation. The PL increase was attributed to trap passivation due to the 08 November 2023 15:51:28 FIG. 14. (1a) PL spectra of a MAPb(Br0.6 I0.4 )3 thin film after sequential cycles of illumination for 2 min followed by 5 min in the dark. (1b) PL emission intensity vs time of a MAPb(Br0.6 I0.4 )3 thin film at the two emission peaks. Reprinted from Hoke et al., Chem. Sci. 6(1), 613–617 (2015). Copyright 2015 Author(s), licensed under a Creative Commons Attribution 4.0 License. (2) Evolution of PL emission monitoring the reversibility of halide segregation when MAPb(Br0.5 I0.5 )3 films are held under four different atmospheric conditions, (2a) vacuum (∼0.2 mbar), (2b) ambient air, (2c) pressurized (2 bars) nitrogen, and (2d) film topped with PMMA and in vacuum. Reprinted from Knight et al., ACS Energy Lett. 4(1), 75–84 (2019). Copyright 2019 Author(s), licensed under a Creative Commons Attribution 4.0 License. (3) LBIC-extracted average EQE (532 nm) of the device during hydration–dehydration cycles. Reprinted with permission from Song et al., Adv. Energy Mater. 6(19), 1600846 (2016). Copyright 2016 John Wiley & Sons. (4a) Dynamic resistance curves of the perovskite film upon exposures to methanol, ethanol, and NH3 at room temperature. (4b) Dynamic resistance curves measured under various moisture levels at room temperature. Reprinted with permission from Hu et al., ACS Appl. Mater. Interfaces 7(45), 25113–25120 (2015). Copyright 2015 American Chemical Society. (5) XRD data on the thermal cycles in air: (5a) In situ thermal cycles in nitrogen environment, two main XRD peak regions. (5b) Reversibility in dry nitrogen. Peak area of the tetragonal 211/103 diagnostic peak (above); peak position and FWHM of the 002/110 peak (below) during in situ thermal cycles (upper panel) in nitrogen environment at atmospheric pressure. (5c) Normalized energy gap trend (E g at 30 ○ C = 1.62 eV). E g has a sudden reduction after annealing at 50 ○ C. The insets represent ideal (left side) vs defective (right side) lattice arrangements. Reprinted with Permission from Alberti et al., Adv. Energy Mater. 9(12), 1803450 (2019). Copyright 2019 John Wiley & Sons. (5d) Comparison on the absorption capability of large-textured and small-random-grained MAPbI3 layers. Reprinted with Permission from Alberti et al., J. Phys. Chem. C 121(25), 13577–13585 (2017). Copyright 2017 American Chemical Society. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-19 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape formation of superoxide species. In contrast, vacuum increases the VII. MULTI STRESSORS trap density probably due to loss of volatile species.55 The authors do Most stability studies of perovskite films, crystals, and full solar note that recovery takes much longer than segregation, something to cells have been carried out in laboratory conditions without follow- bear in mind since the outdoor conditions do not present such long ing a standard protocol and focusing mostly on one main stress relative resting times (see Sec. III). factor. In real outdoor applications, PSC modules will endure simul- Descriptions by Song et al. (see Sec. II) used LBIC EQE map- taneous stressors influencing the device, in varying magnitudes, ping to spatially resolve the loss of EQE during humidity exposure over the course of one day (day/night) and across the year sea- and included two cycles with hydration(80% RH, 90 min) and dehy- sons (weather, sun irradiance, and temperature), or even neutrons dration (dry air, 4 h).26 The EQE decayed from 80% to 20% and in outer space. Stability analyses of PSC under outdoor conditions is completely recovered in dark, as recrystallization of the hydrated the outermost significant assessment because it englobes all the vari- film happened [Fig. 14(3)]. After repeated cycling with longer hydra- ables (at different intensities and doses) that are difficult to mimic tion, some irreversibility was found. It was hypothesized that, once in the laboratory. Outdoor installations are not always feasible for morphological changes occurred, it resulted in a permanent deterio- most of the laboratories and definitively not practical to couple it ration of the photogenerated current. Hu et al. used in situ resistance with most of the in situ advanced characterization techniques; there- measurements to monitor a CH3 NH3 PbI3−x Clx sample conductivity fore, efforts need to be put into replicating outdoor conditions in and the effects produced by different ambient conditions, includ- the laboratory. However, mimicking those exact outdoor condi- ing different solvents for perovskite fabrication [Fig. 14(4a)].25 They tions at laboratory level is a complex task, and real stressors vary tested different RH conditions in nitrogen at room temperature in quantity and intensity and are imposed to the PSC in parallel (20 ○ C) during a cycle going from dry N2 to humid N2 and back or sequentially to each other. They are also randomly imposed on to dry N2 . Figure 14(4b) shows a gradual resistance drop with time the solar cell. Examples of the application of multiple stressor fac- (over 1400 s), directly proportional to the amount of RH. The resis- tors (multi-stress) to PSCs are scarce in the literature, but they are tance quickly recovered to the initial values once humidity was very important, as the outcome of those analysis can shade light removed, independently of the level of humidity. to the working mechanisms taking place under extreme conditions. Alberti et al. studied MAPbI3 thin films (deposited on TiO2 The multi-stress experiment permits to come closer to the operating compact layers) with in situ XRD and in situ spectroscopic ellip- conditions that solar cells should endure and can lead to different sometry (SE) during thermal cycles from 30 to 80 ○ C. Above 50 ○ C, scenarios on the combined effects: 08 November 2023 15:51:28 acceleration of material modification by environmental factors occurred upon transition to cubic lattice, and they proposed (1) Competition: Two or more degradation paths compete and increased accessibility and reactivity of humidity species with the act simultaneously [governed by energy/thermodynamics/ lattice at higher temperatures.105 The authors chose air with 55% saturation (mass transfer)].107–109 RH as a stressor to be below the threshold for hydrate forma- (2) Synergy: One stressor enhances the degrading effect of the tion reported by other works, and indeed no peaks of hydrated other [for example, under heat and light, bias and light, or species were observed by XRD, and all signs were of permanent humidity and light (see Sec. III)].82,110 deformations and configurational disorder in the MAPbI3 (3) Mitigation: One stressor slows down the effect of the other crystal.15,21 When comparing cycles in air (55% RH) and dry (for example, e-beam irradiation diminished illumination nitrogen (0.3 bar overpressure) by SE, it was corroborated that damage,45 applying a bias so that charges are extracted the effects seen were environmentally dependent and irreversible. decrease the illumination damage, or heating the cell causes In situ ellipsometry on samples with different grain size revealed “drying” thus reducing the moisture damage).49,111 that similar phenomena for both small and large grains, supporting (4) The degradation takes different paths (depending on the that the surface stabilization by nitrogen molecules is a universal magnitude of the irradiation, the atmosphere composi- phenomenon. An exception on small grains was seen in an addi- tion or whether the device is at open or short circuit tional healing effect at higher temperatures in nitrogen, and an conditions).89,107,109,112 increase specifically at the lower bandgap absorption was observed (5) Spatial localization: The degradation effect becomes more [see Fig. 14(5)].106 They also saw an increase in PbI2 formation heterogeneous/localized (e.g., at the interfaces with the by in situ XRD under humid air environment compared to N2 selective layers or grain boundaries).108,110 environment, in agreement with other in situ GIWAXS studies.92 Below, we provide a review on the most relevant publica- Despite full recovery being reported throughout repeated tions employing diverse characterization techniques with in situ cycling, it is important to note that these studies use long relative application of various external stressors in various magnitudes and resting time, i.e., the stressing time is much shorter than recov- combinations. Table I summarizes different works classified by the ery. It would be interesting to see more studies making incomplete stress conditions applied. recovery to see the physical accumulative effects of more realistic In Secs. VII A–VII C, we provide detailed explanation of the conditions. An outdoor test would be the ultimate cycling exper- works summarized in Table I, classified by the main stressors applied iment, but to our knowledge, there are no outdoor studies done to the PSC. with complementary in situ characterization of physical mecha- nisms. Nitrogen recovery is used in several examples, and it might be important to point out that nitrogen cycling is not found in any out- A. Illumination and bias door conditions. It is still promising that reversibility possibly can be In situ PL experiments under continuous illumination, open exploited to find a balance for long-term stability. (OC) and short circuit (SC) conditions, and employing different APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-20 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape TABLE I. Summary of the research works performing in situ measurements applying various techniques and combinations of different stressors. T = temperature, RT = room temperature, c- and mp-TiO2 = compact and mesoporous TiO2 , RH = relative humidity, OC and SC = open and short circuit, P = pressure, and UHV = ultrahigh vacuum. ISOS PSC In situ/operando Conditions protocols structure techniques Finding References FTO/c-TiO2 / Pre-accumulated positive ions at the HTL Light mp-TiO2 / interface are redistributed by light-soaking, faster 525/50 nm ISOS-L CH3 NH3 PbI3 / PL (470 nm) at OC bias (drift) slower at SC (diffusion). At longer 110 + OC and SC spiro-OMeTAD at OC and SC light exposure negative ions accumulate at the Au fingers HTL interface under OC bias. Ion migration Au ∼10 nm is slower in larger grains at OC Light mp-TiO2 /Cs0.1 Non-conductive grains produced from halide (422 nm, 0.3 sun) ISOS-L FA0.9 Pb(I0.65 PL (422 nm) segregation due contribute to PL but not to V oc 108 + bias ISOS-V Br0.35 )3 or Cs0.1 + V oc probe because they are isolated. Non-segregation OC and SC FA0.9 PbI3 spiro- (every 3 min) can be distinguished due to coherence in V oc MeOTAD Au calculated from PL and V oc measured Light (white LED Thermal expansion. Excess carriers 100 mW/cm2 ) ISOS-L-3 500 nm XRD and heat from illumination are most 82 + temperature thin film damaging. Switching light on and off at (20, 44, 53 ○ C) CH3 NH3 PbI3 constant temperature has no effect Light (473 nm), FTO/NiOx / Metallic lead formation via PbI2 . Excess 1 W/cm2 ISOS-L-3 Cs:FAPbIx , XPS iodide strongly catalytic toward PbI2 region growth 49 (∼10 suns) I/Pb = 3.8 (rates). Nucleation sites neutralized via metallic T (69 ○ C) and 2.8M lead formation in more stoichiometric samples. Wavelength dependant reaction 08 November 2023 15:51:28 Optical Dry photooxidation and water assisted Various ISOS-D MAPbI3 transmittance photooxidation. H2 O and light synergy for 109 (see Table II) ISOS-L-3 thin film (550 nm) oxidation, strongest at RT. Comparison of rates under different conditions X-ray MAPbI3 Changes dependent on flux. Irradiation ISOS-L-1I films on (NAP-XPS) Degradation path for x-ray exposure. 112 H2 O partial FTO/glass Frenkel defect creation P in UHV (interstitial MA+ and I− ) Light MAPbI3 Production of metallic lead via intermediate (∼AM 1.5 G) ISOS-L 200 nm mp XPS state (CH3 NH3 )1−x PbI3−x to PbI2 to metallic lead 111 + vacuum ISOS-V TiO2 / and iodide. Photogenerated carriers induce + bias (OC/±1 V) spiro/Au chemical and optoelectronic change if not extracted Light/ Vacuum chamber Competing reactions. CH3 NH2 and HI from a Temperature ISOS-L-2 FAPbI3 with mass reversible reaction and CH3 I and NH3 from an 89 and Vacuum or T ISOS-LC powder spectrometer irreversible reaction. PbI2 degradation into Pb0 107 Vacuum MAPbI3 (MS) and I2 (gas). Light and temperature affect balance (in cycles) powder between reversibility. FAPbI3 is more complex Light Single crystal Conductivity variations in test device Temperature ISOS-L-2 MAPbBr3 EIS, depending on environmental pressure, 113 (70% RH) ISOS-V with contacts G-KPFM resistance increase with increase in humidity Bias while the device was under bias in the dark. Triple phase boundary at interfaces perovskite composition revealed distinct conclusions from different probing were done on the bulk thickness rather than the interface.108 laboratories. Ebadi et al. used a combination of PL and electrolumi- They studied two Cs-FA cation perovskite, one with only iodide and nescence (EL) to investigate the origins of V oc and J sc loss under light a mixed I–Br halide that undergoes the known halide segregation soaking. In their work, a 422 nm homogeneous illumination with a [see Sec. III (illumination)]. The cell was switched from OC to SC diffused laser (30% sun) was used to ensure that illumination and while recording “in operando” the transients of the J sc and V oc and APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-21 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 15. (1a) Evolution of open-circuit voltage (V oc ) and PL intensity during light soaking of a device containing CsFAPbI3 . (1b)–(1f) Light-soaking experiment of a device containing Cs0.1 FA0.9 Pb(I0.65 Br0.35 )3 for 10 000 s. In region I, the device is mainly kept at short circuit and periodically (every 3 min) switched to open circuit. (1b) V oc , (1c) current density, (1d) energy of the PL maximum, and (1e) counts at the PL maximum. (1f) Samples of spectra during region I, under short circuit (black) and the subsequent open circuit conditions (red). Reproduced with permission from Ebadi et al., J. Mater. Chem. A 9(24), 13967–13978 (2021). Copyright 2021 from The Royal Society of Chemistry. (2a) Schematic diagrams of sample structure and metallization pattern. (2b) Schematic diagram of PL characterization method. Reproduced with permission from Deng et al., Nano Energy 46, 356–364 (2018). Copyright 2018 Elsevier. (3a) Scheme of the in situ XPS setup on a complete solar cell device, probed near the top electrode. (3b) Surface elemental ratio of MAPbI3 under light over the time in vacuum. (3c) The change in elemental ratio of the MAPbI3 surface determined from core level spectra over time. Reprinted with permission from Das et al., Phys. Chem. Chem. Phys. 20(25), 17180–17187 (2018). Copyright 2018 from The Royal Society of Chemistry. its corresponding PL intensity and emission wavelength. Both V oc migration was influenced by other effects, such as photobrighten- and PL intensity decreased under illumination, and a good corre- ing [Fig. 15(1e)]. When investigating Cs0.1 FA0.9 Pb(I0.9 Br0.1 )3 and lation between the calculated V oc from the PL yield and the actual Cs0.1 FA0.9 PbI3 , the mixed halide showed halide segregation above V oc was found [Fig. 15(1a)]. The effect shown by this relation was certain illumination thresholds, and the behavior of mixed halide hypothesized to be caused by a build-up of non-radiative recombi- perovskite was indeed more complex. Upon light soaking under nation at the perovskite interfaces to spiro-OMeTAD and mp-TiO2 , SC for the mixed halide, with interludes to check V oc , a PL red i.e., changes in V oc depend on the non-radiative recombination at shift (720–770 nm) was observed [Figs. 15(1d) and 15(1f)], corre- the contacts via a barrier or traps, yet the physical origins were not sponding to a I:Br ratio modification from 0.65:0.35 to 0.8:0.2. In discussed. Interestingly, the PL yield of compositions prone to halide addition, they observed that the reduction of J sc with light soaking APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-22 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape [Fig. 15(1c)] related to the formation of isolated highly emissive boundaries.110 It is worth mentioning that this PL characteriza- perovskite clusters in the film, thus not contributing to the current tion was done in air and that bias application was interrupted by output of the device. These regions originate from halide segregation periodic j–V measurements to complement with the device perfor- to small bandgap iodine-rich clusters. Interestingly, single halide mance along the experiment. It is probable that the bias perturbation perovskites (MAPbI3 and MAPbBr3 ) had similar behavior of PL and moisture may have influenced the actual device evolution. increase, the authors hypothesized that ion migration could create Future works would benefit to include a control sample without regions that cannot extract photocarriers. The formation of these perturbations to rule out its effect in the experiment. clusters does not seem to form recombination “traps” or “defects” The dependence of the applied bias on the device on the degra- since the V oc values remain constant along 6000 s of J sc − V oc dation upon light soaking was also studied by Das et al. via XPS cycling while the emission had a progressive redshift and intensity to reveal the differences in chemical composition.111 MAPbI3 -based increase. Such finding challenges the state-of-the-art knowledge cor- complete PSC devices were investigated in situ by XPS under contin- relating high PL to high V oc , in contrast to the findings described uous illumination or dark and near to a biased electrode measured earlier for this same publication. The authors propose that whereas 2 mm-away from an electrode where bias was applied (+1 or 1 V) the “high PL = high V oc ” assumption [Fig. 15(1a)] is true for flat [Fig. 15(3a)]. XPS under light soaking (spectra close to AM 1.5 G, quasi-Fermi levels (q-EF ), the ion mobility in perovskite can create 100 mW/cm2 ) at OC showed a gradual decrease in the N1s sig- gradient q-EF .108 nal [Fig. 15(3b)], caused by the removal of volatile organic species Other authors have reached different conclusions regarding driven by the ultra-high vacuum (10−9 Torr), and the appearance of how recombination processes vary under illumination at different Pb0 core level signals together with new states in valence band (VB) biases via PL detection. Deng et al. studied the effects of con- region. In contrast, applying a bias of +1 V, below V oc (thus extract- tinuous light soaking (525/50 nm bandpass filter) on a complete ing photogenerated carriers) was found to be beneficial to slow down device, targeting the spiro-OMeTAD/MAPbI3 interface region by the decomposition speed [Fig. 15(3c)], while applied bias in the dark means of steady state and time-resolved photoluminescence (PL and TRPL).110 The light (∼0.72 sun) shone through the glass side as in did not induce any visible changes in their experiment. With this, regular operation of a solar cell, while a confocal laser microscope they demonstrate that photo-instability is intrinsic of the perovskite focused close to the cathode electrodes, capturing the PL emission rather than induced by ambient O2 or water. It is worth mentioning (470 nm laser excitation) through semi-transparent 10 nm top Au that, due to the XPS surface sensitivity limitation, it is not possible to monitor the actual biased perovskite under the contact.111 08 November 2023 15:51:28 layer between thicker finger electrodes 2 mm-apart [Figs. 15(2a) and 15(2b)]. This method allowed for probing the HTM inter- face by PL in a fully operative device (n-i-p).110 It was found that the PL and TRPL evolved along 10 min illumination both at OC and SC; however, slight differences were found between OC and B. Illumination and heating SC. Changes were attributed to the internal redistribution of ions. While in situ XPS characterization can infer the causes of Initially, illumination is believed to redistribute pre-accumulated perovskite degradation by observing, for example, a reduction of cations at the HTM interface back into the film bulk, a process partly N1s signal or increase in metallic lead, the exact released com- reversible in the dark. An initial PL increase at OC was observed pounds are elusive to this technique. In this regard, Juarez-Perez due to reduced nonradiative recombination and PL reduction at SC et al. detected in situ the composition of the volatile degradation due to better charge extraction upon annihilation of the ions that molecules of PbI2 , MAPbI3 , and MAPbBr3 powders under illumina- pre-accumulated in the dark. The main difference was that, with tion and heating cycles by coupling a mass spectrometer (MS) to a longer illumination, at OC, the PL intensity and lifetime decreased customized vacuum chamber [Figs. 16(1a)–16(1e)].107 Upon merely substantially, while at SC, the PL properties remained constant. exposing the MAPbI3 powders to vacuum, the MS started detect- PL decrease at OC was hypothesized to be caused by the internal ing signals of the decomposition species, including I2 , CH3 NH2 , HI, built-in field that produces a drift of negative ions to the HTM CH3 I, and NH3 , and occluded dimethylformamide (DMF) solvent interface, creating traps and thus enhancing recombination. At SC, molecules, indicating the low stability of such material. Illumination the ion movement is mainly through diffusion, and, without con- and temperature cycles of 5 min resulted on an overall increase of centration gradient, accumulative ion motion has no driving force. all these signals [Figs. 16(1a) and 16(1b)]. Interestingly, they found A particularity of this in situ study was the very short dwell time that, once illuminated with 0.5 suns, MAPbI3 continues to decom- (2 ms) used to scan the sample, allowing to capture the instanta- pose in the dark, even after several minutes. Similarly, photolysis neous PL without build-up effect (and interface recombination from of PbI2 by releasing I2 was also found, in all cases leaving Pb0 in the TiO2 -perovskite interface) up to a depth of ∼50 nm from the the remaining powders, in accordance to previous findings.42,111 perovskite/spiro-MeOTAD interface. Through the confocal scan- Photodecomposition of PbI2 to I2 was found to have a lower acti- ning configuration, an increase in dynamics at clusters with smaller vation energy (Ea ) than that of illuminated MAPbI3 ; thus, any grains was observed, contributed to higher boundary/surface to vol- impurities existing in the film may initiate further decomposition ume ratio in these grains. Both the increased pre-accumulated ion reactions.38,43 In addition, heating PbI2 in the dark did not trigger concentration and increased ion diffusivity and concentration at I2 release, but it did for heating MAPbI3 , indicating that MAPbI3 is grain boundaries may have sped up migration. With ex situ 2D an unstable perovskite in general. MAPbBr3 was found to be more PL mapping, the authors demonstrated a more uniform distribu- stable than its iodine counterpart because it released only CH3 NH2 tion of PL intensities at SC, while at OC, smaller grains showcased and HBr under illumination and/or mild heating (40–80 ○ C). faster rates in changes, indicating increased ionic transport at grain CH3 NH2 and HX (X = halide) are reversible decomposition APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-23 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 16. (1) Mass spectrometry profiles of MAPbI3 decomposition products during illumination and heating-in-the-dark pulse experiments. (1a) Light/dark intervals (5 min each) on MAPbI3 perovskite using a Xe-lamp delivering 55 mW/cm2 of light power. (1b) Heating on/off intervals (5 min each) on the MAPbI3 sample under dark conditions. Species of interest detected in MS are labeled on the right side. The right panel shows calibrated mass traces for (1c) CH3 NH2 , (1d) CH3 I, and (1e) I2 during the heating intervals under dark conditions. Reprinted from Juarez-Perez et al., J. Mater. Chem. A 6(20), 9604–9612 (2018). Copyright 2018 Author(s), licensed under a Creative Commons Attribution 4.0 License. (2) Time-resolved in situ XPS Pb 4f areal fit results from Pb/I = 3.2 (2a) and Pb/I = 2.8 (2b) perovskite samples. Molar fraction of Pb is normalized by the total Pb 4f signal. Reprinted with permission from Donakowski et al., ACS Energy Lett. 6(2), 574–580 (2021). Copyright 2021 American Chemical Society. (3a) Change of temperature and strain with illumination duration time and relation of light-induced strain and sample temperature for in CH3 NH3 PbI3 single crystal under illumination of 100 mW/cm2 white LED. (3b) XRD patterns of CH3 NH3 PbI3 single crystal under light off at room temperature (RT), light (100 mW/cm2 , white LED) on at 53 ○ C, and light off at 53 ○ C. (3c) XRD patterns of CH3 NH3 PbI3 single crystal powders under light off at RT and under light on at 44 ○ C. Reprinted with permission from Chen et al., Adv. Mater. 31(35), 1902413 (2019). Copyright 2019 John Wiley & Sons. products that convert back to perovskite spontaneously, in con- above 95 ○ C, the HCN dimer fragment, HCN–(H)CN attributed trast of the irreversibility of CH3 X and NH3 . Similar conclusions to the presence of sym-triazine.89 There is an agreement in these were obtained in a later publication for FA-based perovskites.89 in situ works that the illumination is an important cause of in- Upon inserting FAPbBr3 perovskite into vacuum, MS detected sig- stability for the MAPbI3 , releasing MA and halide. Changing the nals of the degasification products FA, HCN, and NH3 , which did halide to Br increases its stability, and substituting MA with the not increase with mild illumination (0.55 suns, <60 ○ C). Heating longer alkyl cation FA increases its light and temperature resis- over 75 ○ C in the dark produced a significant increase in the above- tance.89 Regarding the relationship between compositional ratio and mentioned signals, and the appearance of HBr, Br isotopes and, stability, Donakowski et al. demonstrated the utter importance of the APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-24 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape halide excess/deficiency of (Cs,FA)PbI3 in the kinetics and mech- of a contacted single crystal perovskite.113 Interaction of the per- anism during thermal and photo-decomposition via in situ XPS.49 ovskite with the environment can be avoided by encapsulation or By fitting the different peaks of the Pb 4f5/2 core-level spectra, the passivation but, instead, interaction with its own degradation prod- authors deconvoluted the signals of Pb0 , PbI2 , and FAPbI3 . When ucts should be considered.109 Albeit the environment alone seldom conditioning Cs:FAPbI3 thin films (FTO/nickel oxide/perovskite) produce drastic changes on perovskites (unless in extreme con- at 69 ○ C and 473 nm light irradiation (1000 mW/cm2 , over a cir- ditions), Siegler et al. prove that the atmosphere gases can have cular area of 800 μm diameter), excess I− is shown to catalyze drastic impact under operational conditions. Optical transmittance the production of metallic lead via PbI2 impurities. While near- of 550 nm light was used to quantify degradation rates under stoichiometric samples (Pb/I = 2.8) showcased only a small increase various combinations of stressors (see Table II) and illumination of Pb0 signals [Fig. 16(2a)], excess iodide samples (Pb/I = 3.8) intensities. By adapting the probe and detection wavelengths to showed a rapid loss of FAPbI3 simultaneous to PbI2 formation, exclude the absorption/emission of the solid degradation products, followed by a gradual formation of Pb0 , suggesting a two-step they could monitor exactly the disappearance of MAPbI3 in a thin reaction [Fig. 16(2b)].49 film over time in situ. Two degradation pathways could be distin- It is very common to find discrepancies in the literature regard- guished: (i) a dry photo-oxidation induced by the interaction of ing the beneficial or detrimental effects of illumination on per- photoexcited electrons with the O2 and (ii) a humidity accelerated ovskites. Motti et al. demonstrated that such conflicting reports can photo-oxidation. originate from the competing processes happening simultaneously The increase in degradation rate with water and oxygen could in the perovskite film, which, depending on the sample conditions, not be explained by additive effects but only by synergy between one can dominate over other processes.114 To demonstrate that, water and oxygen in an accelerated mechanism. It was proposed they employed in situ PL measurements with combined light and that H2 O lowered the kinetic barrier for photo-oxidation as well thermal stress. Illumination was tuned to various wavelengths and as providing additional pathways to form reactive oxygen species, laser pulse repetition rates (varying incident power) and the sam- the superoxides are considered the rate-determining step for photo- ple was heated in inert atmosphere. It was concluded that the long oxidation. We note that, in this work, dark conditions could not be living trapped carriers mediate the ionic movement under illumi- considered in situ since the probe changes the conditions by illumi- nation. This induces different photochemical reactions that cause nation, something supported by the rate in dark being affected by the either PL increase or decrease. Moreover, their work demonstrates sampling frequency when oxygen was present. In inert atmosphere, 08 November 2023 15:51:28 that passivating the coordinated I2 by organic additives (or other accelerated degradation was negligible either by probe laser or keep- passivation strategies) is key for a substantial enhancement of the ing the light on or off. Mass transport limited and kinetically limited device stability. processes are discussed with regard to the presence of water dur- ing degradation. These experiments were performed in high vacuum under gas flow to ensure kinetically limited conditions not slowed by mass transport. In real operational systems, mass transport may C. Illumination, bias, and temperature become dominant, especially at longer timescales. Water and oxy- Elevated light-induced heat and excess of photogenerated car- gen may need to compete for adsorption sites on the surface, and riers were also found to be the main culprits in rapid photoinduced perovskite degradation. Chen et al. studied the effects of various light-induced stresses (heat, lattice strain, electric field, and charge TABLE II. Observed degradation rates for all possible combinations of environmental carriers) in the stability of MAPbI3 .82 The light-induced lattice strain stresses. was tracked with in situ XRD while the temperature of the sample Observed rate of MAPbI3 disappearance (mol m−2 s−1 ) was recorded [Fig. 16(3a)]. Upon illumination with 100 mW/cm2 , a shift to lower 2θ was observed uniformly across the reported XRD Environmental condition 25 ○ C 85 ○ C peaks, indicating an increase in the lattice strain. With further mea- surements with heating in the dark [Fig. 16(3b)], it was found that H2 O, O2 , lighta 1 × 10−7 2 × 10−7 the strain was mostly induced by the light-induced heat. By mea- O2 , light 2 × 10−9 1 × 10−7 suring powder XRD, it was found that the light induced strain H2 O, light 5 × 10−10 2 × 10−9 (T = 40 ○ C) was orientation-dependent [Fig. 16(3c)], in contrast H2 O, O2 , dark <1 × 10−11 2 × 10−9b with previous reports. With interferometry measurements, sub- H2 O, dark <1 × 10−11 5 × 10−11 millisecond changes in the film thickness could be detected, con- O2 , dark 2 × 10−10b 1 × 10−9b firming that neither illumination alone nor piezoelectricity con- Light, inert <1 × 10−11 2 × 10−10 tributed to the lattice strain. However, the rapid initial degradation Dark, inert <1 × 10−11 3 × 10−10 of PSC was not due to such lattice strain but originated by the cou- pled effect of heating and excess carriers that catalyze the process of a H2 O: 50% RH; O2 : air (21% oxygen); light: 1 sun equivalent photon flux; inert: pure defect formation.82 N2 (<0.01% O2 ). b Another important implication of approaching operational Samples in dark oxygen-containing environments degrade through photo-oxidation while the probe beam is active, making it unclear whether a distinct oxidation path- conditions during in situ characterization is considering the envi- way operates under fully dark conditions (extended discussion can be found in the ronmental factors of PSC, independently on whether cells are encap- supplementary material of Ref. 109). Reprinted with permission from Siegler et al., sulated or not. It has been shown that the environmental conditions J. Am. Chem. Soc. 144(12), 5552–5561 (2022). Copyright 2022 American Chemical and variation of gas-pressure show an impact on the conductivity Society. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-25 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape passivation layers may occur on the surface from degradation prod- device stability under the compounded stress of temperature, light, ucts. If any water is created during any degradation in encapsulated bias, humidity, and environmental stress cycles. samples, it could have detrimental effects on stability.109 The exper- Velilla et al. reported in 2021 that while 19 407 documents were iments done by Siegler et al. demonstrated that, depending on the found when searching for “perovskite and solar cell” on SCOPUS environmental conditions, perovskite degradation can take different (cite) database, only 100 articles could be found when adding the pathways.109 Related to this, changes in the degradation mechanisms word “outdoor.”121,122 Köbler and Khenkin et al. have tabulated can also be induced in more extreme conditions, such as high energy 14 outdoor stability studies (ISOS-O) on single junction perovskite photon flux. To prove that, Kot et al. employed in situ near-ambient- from 2015 to 2021. They noted that in all the studies (except one) the pressure x-ray photo-electron spectroscopy (NAP-XPS) at almost device is set at open circuit while periodic j–V curves are measured dark conditions with very low intensity x-ray exposure (6.1 ⋅ 109 in the solar simulator (ISOS-O-1) or under real sunlight (ISOS-O-2), photons/s) and different H2 O vapor pressure (10−3 , 10−2 , 10−1 , and not specified in the table. One of the works employed ISOS-O-3,123 1 mbar).112 No significant changes of the XPS spectra over time were where the MPP is monitored in situ, which is the closest condition obtained in dark, independently of the water vapor pressure. How- to in operando conditions and the recommended approach to per- ever, high x-ray photon fluxes (7.4–9.9 ⋅ 1011 photons/s, 520 eV) in form the measurement. Previous studies have shown that periodic humid conditions induced a pronounced degradation with changes OC/J–V measurements provide different stability results than MPP in all the core level element binding energies. The authors pro- tracking, as the ionic migration at different internal bias of the cell pose that irreversible chemical reactions between the perovskite largely influences the degradation mechanisms.124 and water are catalyzed by the x-rays, in agreement with Das Reports on the prediction of PSC lifetime from stability studies et al.111 The irreversible decomposition from the break of the C–N have shown to be distinct depending on the type of analysis: indoor bond generated a new N1s peak, attributed to NH3 , an irreversible or outdoor. Currently, several important projects are being estab- decomposition product also proposed by Juarez-Perez et al.107 The lished where indoor and outdoor stability analyses are combined changes were attributed to the creation of Frenkel defects (such with machine learning tools. Thus, a surge of interesting new results as interstitial MA+ and I− interstitials). It was proposed that this is expected soon. However, some interesting relationships have been mechanism was mainly due to the weakening of the structure due initially observed and, in a few cases, similar results were reported, as to humidity, together with the intense radiation, yet oxygen pres- will be described below. There are some aspects of outdoor charac- ence could not be entirely excluded because of OH− bonds created terization that are difficult to incorporate into predictions of PSC 08 November 2023 15:51:28 during sample transportation and as the method is highly surface stability performance from indoor testing. Indoor stability setups sensitive.112 have the advantage of applying well-controlled dosage-specific com- bination of stressors with advanced in situ material characterization, while outdoor conditions impose multiple stressors in a variable and VIII. TOWARD THE PREDICTION OF OPERATIONAL unpredictable way. STABILITY OF PSC UNDER OUTDOOR CONDITIONS Song and Aernouts argued the importance of meteorological Before the publication in 2020 of the ISOS protocols for the and diurnal variation during outdoor testing and indicated that stability assessment and reporting of PSCs,6 and even in later pub- variable conditions are essential if indoor testing must reflect real lications, the reports of PSC stabilities were difficult to compare stability.125 A PSC that reaches a steady state under accelerated test- and unify in a table or chart.6 A few review articles have attempted ing conditions or during a constant stress test does not necessarily to tabulate the record stabilities reported including the most rele- show a stable behavior outdoors. The large variation on diversity vant details on the stress conditions and measurement protocols, and intensity of outdoor stressors, such as temperature and light, which largely varies among publications. Noticeable is the lack occurs in diurnal and seasonal cycles. Therefore, it is important to of information in the description of measurement conditions in emphasize that, although accelerated testing can prove the resistance many publications reporting stability, where we find missing tem- of a device against certain conditions, resistance against variation of perature or humidity values, incomplete characterization protocol multiple stressors in a rather unpredicted manner would be the end description, or normalized stability without the value of the PCEs goal for a solar cell with high lifespan.125 reported.115–119 In this section, we will provide a brief overview of There are some reports successfully reproducing indoor studies the state-of-the-art of outdoor studies, as an effort to correlate out- in outdoor testing. Emery et al. compared PSC fabricated with two door and indoor measurements and machine learning approaches to different encapsulation methods and could, by the ISOS-D3 (IEC predict stability. 61215, 1000 h at 85 ○ C and 85% RH) protocol, see how results in the Zhang et al. have provided the first study employing a statistical indoor controlled environment corresponded to the outdoor results comparison of a large stability data obtained from the reported lit- [see Fig. 17(3)].126 Devices that retained over 95% PCE at ISOS-D3 erature and gathered in the Perovskite Database.12,13,120 Over 7000 also passed an outdoor stability test according to ISOS-O3 proce- stability reports were categorized into the closest ISOS protocols dures for over 10 months. The authors attempted to diagnose failure [see Fig. 17(2)], and the authors proposed a unified indicator to mechanisms in devices that did not pass the test, but unfortunately, enable the comparison between different stress conditions that led ex situ EL and JV measurements were not sufficient to reveal what to several generalized conclusions. The authors emphasize that more microstructural, compositional, and interfacial changes occurred in detailed data are needed to feed the algorithms and extract more devices when changes in performance were seen.126 solid conclusions in the future. Their study is a great example of why Köbler et al. showed that, for two different device structures, data need to be standardized and why it is important to discuss prac- indoor constant illumination tests failed to predict the outdoor tice of measurement, including more accurate models for describing behavior of the PSCs, but cycling test ISOS-LC1-I could predict APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-26 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 08 November 2023 15:51:28 FIG. 17. (1) Patterns of degradation in the maximum power and the ideality factor. (1a)–(1c) Maximum power under different power rating conditions suggested by IEC 61863 for the convex (1a), linear (1b), and concave (1c) degradation behavior patterns. (1d)–(1f) Ideality factor (nID) analysis of the same samples: (1d) exhibits a concave pattern for samples exhibiting a convex Pmax pattern, (1e) nID exhibits a linear pattern for samples exhibiting a linear Pmax pattern, and (1f) nID exhibits a convex pattern for samples exhibiting a concave Pmax pattern. The thick black line on each box represents the average performance in the corresponding time window. Adapted with permission from Velilla et al., Nat. Energy 6(1), 54–62 (2021). Copyright 2021 Nature Publishing Group. (2) The distribution of stability protocols used for stability data in the Perovskite Database. The most widely used protocol is ISOS-D-1, where devices are stored in ambient air in the dark at open circuit. Adapted from Zhang et al., Nat. Commun. 13(1), 7639 (2022). Copyright 2022 Author(s), licensed under a Creative Commons Attribution 4.0 License. (3a) Evolution of PSC performance ratio of glued (“LAB”) and laminated (“COM”) cells during the outdoor exposure. Days of indoor control measurements are shown in gray. (3b) Irradiance in the plane of cells and cell temperature during the outdoor experiment. Adapted from Emery et al., ACS Appl. Mater. Interfaces 14(4), 5159–5167 (2022). Copyright 2022 Author(s), licensed under a Creative Commons Attribution 4.0 License. (4) Comparison of indoor and outdoor stability for two solar cells using a different HTL. (4a) Indoor experiment with constant illumination. (4b) Indoor experiment with cycled illumination. Both indoor experiments are conducted at 25 ○ C and in nitrogen. (4c) Outdoor experiment, conducted in Berlin, July–August 2020. Values not averaged for better display; averaged performance ratio over five and six cells. All experiments (a)–(c) were performed under continuous MPP-tracking. Adapted from Köbler et al., “The challenge of designing accelerated indoor tests to predict the outdoor lifetime of perovskite solar cells” (2021). Copyright 2021 Author(s), licensed under a Creative Commons Attribution 4.0 License. outdoor performance for one of the devices.127 When aging tests day. Higher performance from irradiance soaking and higher tem- were performed following ISOS-L1I [constant illumination, 25 ○ C, perature both peak at midday, resulting in an apparent positive effect indoor, Fig. 17(4a)], ISOS-LC-1I [cycled illumination, 25 ○ C, indoor, of the temperature on device efficiency, something contested by Fig. 17(4b)], and ISOS-O2 [outdoor, Fig. 17(4c)], the different kinet- controlled studies under indoor conditions. These common trends ics of the slow transients between samples delivered different results propose a problem in outdoor data analysis, making in situ outdoor when comparing ISOS-L1I, ISOS-LC-1I, and ISOS-O2. The authors measurements a necessary tool for PSC commercialization. Still, this point out the importance of exercising caution when correlating also argues the importance of having performed indoor studies to stress changes to photovoltaic parameters during outdoor measure- achieve the separate picture of stressor effects before trying to see ments, as levels of stress factors often change together during the the effects of a multitude of stressors varying simultaneously. APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-27 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape In tandem devices, the comparisons of silicon SCs and PSCs authors working on big data analysis, indicating that research efforts have led to observations in what outdoors variations set off a differ- may be inconsistent due to differences in sample quality and history, ence PSC stability compared to its silicon counterpart. Babics et al. common practices, laboratory facilities, among others. Despite the identified that light intensity and temperature variations in combi- current challenge to categorize the reports into a certain ISOS proto- nation with dust accumulation (soiling) differentiate outdoor from col, it is encouraging that recent publications provide more complete indoor laboratory conditions when monitoring all the photovoltaic description of the methods, indicating a general trend on reaching parameters and matching current between the Si and perovskite part standardized stability measurements. This is particularly interesting of tandem device.128 to facilitate machine learning studies to make predictions and design Tress et al. performed long-term indoor operational tests on and validate the accelerated indoor tests. efficient PSC devices under a simulation of temperature and illu- mination outdoor conditions, flushed with dry nitrogen, to find IX. GOOD PRACTICES AND CONCLUSIONS reversible degradation as the main responsible factor for efficiency fluctuation. They observed that the intrinsic response-differences of In situ and operando characterization of PSCs under different the PSC in these conditions were dependent of the time of day and stressors has provided significant insights into physical, chemical, season. Changing conditions, as outlined in Secs. VI and VII, can mechanical, and electrochemical transformation in these materi- produce synergetic, competitive, escalating effects.129 als and systems. Given the “soft material” nature of halide per- By monitoring the temporal evolution of the MPP and JV ovskites, where reversible effects and special features are observed at curves of PSC minimodules tested at an outdoor testing site, short timescales (e.g., recovery in the dark, ion migration), in situ Velilla et al. observed three different trends in MPP loss vs and operando characterization appears as the holy grail on pro- time (degradation-shape): concave, linear, and convex evolution viding new degradation insights from continuous monitoring. We [Fig. 17(1)].121,122 The mean solar cell junction temperature as a have observed that most characterization techniques investigate function of solar irradiation, ambient temperature and thermal singular physics/chemistry of the material system; however, there properties of the photovoltaic material, and IV curves from tested is strong interdependent interplay of degradation mechanisms in devices obtained every minute helped the authors gain suspicion of PSCs, which can only be identified under in situ and operando the physical mechanisms that could be responsible for the different conditions. slopes in the power decrease. By monitoring the ideality factor evo- These methods can accurately monitor the relationship 08 November 2023 15:51:28 lution, the authors added another measure of the quantitative cell between different stressors imposed sequentially or at one single degradation, one that combined readout from cells that can be more time, resembling the real testing conditions of a PSC. The applica- related to the device physical and chemical properties, although tion of in situ multi-modal characterization under several stressors physical phenomena were not immediately observed. has not been reported yet but is foreseen as the optimal step in Zhao et al. observed that the degradation of PSC followed an characterization complexity. Arrhenius dependence on temperatures between 35 and 100 ○ C in It is apparent from the publications mentioned in this article their experiments at constant illumination. This observation was that many in situ studies have multi-stressor components, although used to predict the 5-year lifetimes for a certain cell stack. Yet they multi-stressor comparison might not have been the main target of assumed continuous temperature of 35 ○ C for this prediction, which the studies. The source of stressor, especially for the light/thermal is far from the actual outdoor operating conditions, and recent stressors section, describes an ambiguity as one stressor may be the studies have debated the validity of constant condition tests to por- source of the other, unless conditions are carefully controlled. A sim- trait real operational stability with outdoor unpredictable conditions ilar response is observed for environmental conditions where the under variation.10 removal of gases has indeed been studied as a separate environ- While accelerated testing may provide a shortcut to seeing the ment. Different environments should be considered in cases where processes that occur over time, quantification of the performance the atmosphere is in continuous flow mode (such as in a holder lifetime of a device by prediction from accelerated tests has not yet with N2 flux) or the device is sealed, and the initial volume is the been demonstrated. Reproducing exact outdoor conditions along same throughout experiments. This is especially important since operation lifetimes of years in indoor setups is extremely challenging encapsulation is crucial for the commercialization of PSCs, and thus, and unpractical, limiting the accuracy of the conclusions extracted. analyses under gas flux is probably not the most accurate condition To establish a correlation between indoor and outdoor stability mea- to understand PSCs degradation. surements together with material changes, the scientific community Many are the working mechanisms behind PSC degradation must work in parallel toward (i) increasing the reports on outdoor that are still unresolved, for example, localized heating and hotspot stability tests for various device structures and perovskite composi- phenomena, and the effect of irradiation under vacuum or extreme tions, together with additional in situ characterization besides j–V or temperature (e.g., freezing) cycling. Similarly, the nature of reverse MPP tracking and (iii) increasing the parameter combination of the bias voltage imposed to PSC (related to shading under real operation simulated stressors that are applied in indoor setups imitating the conditions) and its effect on irreversible degradation and hysteresis ambient variations. It is worth mentioning that correlating indoor are still open questions that need to be resolved. and outdoor tests and reaching a consensus on its relationship and We expect that in situ and operando characterization will be the development of accelerated tests seems only feasible through the extrapolated to the scale up and fabrication of PSCs in a production implementation of automated analysis and machine learning that line (from synthesis, processing, and integration of the components can cope with the vast parameter space, statistics, and weather vari- at industrial level). Although consensus protocols for stability assess- ability. It is also important to consider the conclusions obtained by ment of PSC have already been reported, these are still missing for APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-28 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape the characterization of PSC technology. In this regard, we must also REFERENCES consider that the current technological benchmark for PSCs com- 1 W. Shockley and H. J. Queisser, “Detailed balance limit of efficiency of p-n prises modeling approaches and relies on AI tools, such as machine junction solar cells,” J. Appl. Phys. 32(3), 510–519 (1961). learning. Therefore, it is crucial for the practical implementation of 2 J. Park, J. Kim, H.-S. Yun, M. J. Paik, E. Noh, H. J. Mun, M. G. Kim, T. J. Shin, and PSC technology to move the attention onto technological aspects S. I. Seok, “Controlled growth of perovskite layers with volatile alkyl ammonium related to in situ and operando characterization under real-world chlorides,” Nature (published online). 3 functioning of the entire PV systems. National Renewable Energy Laboratory, Best Research-Cell Efficiency Chart, 2023, https://www.nrel.gov/pv/assets/pdfs/best-research-cell-efficiencies. pdf. 4 M. H. Futscher and B. Ehrler, “Modeling the performance limitations and ACKNOWLEDGMENTS prospects of perovskite/Si tandem solar cells under realistic operating conditions,” ACS Energy Lett. 2(9), 2089–2095 (2017). We thank the Project ProperPotoMile, which is supported 5 M. O. Reese, S. A. Gevorgyan, M. Jørgensen, E. Bundgaard, S. R. Kurtz, D. S. under the umbrella of SOLAR-ERA.NET Cofund 2 by the Spanish Ginley, D. C. Olson, M. T. Lloyd, P. Morvillo, E. A. Katz et al., “Consensus stabil- Ministry of Science and Education and the AEI under the Project ity testing protocols for organic photovoltaic materials and devices,” Sol. Energy No. PCI2020-112185 and CDTI Project No. IDI-20210171. The Mater. Sol. Cells 95(5), 1253–1267 (2011). 6 ICN2 was supported by the Severo Ochoa Centers of Excellence M. V. Khenkin, E. A. Katz, A. Abate, G. Bardizza, J. J. Berry, C. Brabec, F. program (AEI) (SEV-2017) and is currently supported by the Brunetti, V. Bulović, Q. Burlingame, A. Di Carlo et al., “Consensus statement Severo Ochoa Centers of Excellence program (Grant No. CEX2021- for stability assessment and reporting for perovskite photovoltaics based on ISOS procedures,” Nat. Energy 5(1), 35–49 (2020). 001214-S), both funded by MCIN/AEI/10.13039.501100011033, and 7 Y. Li, B. Shi, Q. Xu, L. Yan, N. Ren, Y. Chen, W. Han, Q. Huang, Y. Zhao, and also funded by the CERCA Program/Generalitat de Catalunya. X. Zhang, “Wide bandgap interface layer induced stabilized perovskite/silicon tan- Thanks to the Agència de Gestiód’Ajuts Universitaris i de Recerca dem solar cells with stability over ten thousand hours,” Adv. Energy Mater. 11(48), (AGAUR) for the support to the consolidated Catalonia research 2102046 (2021). group 2021 (Grant No. SGR 01617) and the Xarxa d’ R + D 8 G. Grancini, C. Roldán-Carmona, I. Zimmermann, E. Mosconi, X. Lee, D. Mar- + I Energy for Society (XRE4S). S.R.R. acknowledges the sup- tineau, S. Narbey, F. Oswald, F. De Angelis, M. Graetzel, and M. K. Nazeeruddin, port from “la Caixa” Foundation (ID 100010434) with fellowship “One-year stable perovskite solar cells by 2D/3D interface engineering,” Nat. code LCF/BQ/PI20/11760024 and Grant No. PID2021-122349OA- Commun. 8, 15684 (2017). 9 P. Guo, H. Zhu, W. Zhao, C. Liu, L. Zhu, Q. Ye, N. Jia, H. Wang, X. Zhang, W. 08 November 2023 15:51:28 I00 funded by MCIN/AEI/10.13039/501100011033 and “ERDF A Huang et al., “Interfacial embedding of laser-manufactured fluorinated gold clus- way of making Europe.” The authors acknowledge the Spanish Min- ters enabling stable perovskite solar cells with efficiency over 24%,” Adv. Mater. istry of Science and Innovation for the predoctoral contract to F.B. 33(36), 2101590 (2021). with Reference No. PRE2020-092669 of the Project No. SEV-2017- 10 X. Zhao, T. Liu, Q. C. Burlingame, T. Liu, R. Holley, G. Cheng, N. Yao, F. Gao, 0706-20-3. The authors thank Dámaso Torres for his valuable help and Y.-L. Loo, “Accelerated aging of all-inorganic, interface-stabilized perovskite with 3D image design and rendering. solar cells,” Science 377(6603), 307–310 (2022). 11 S. Sun, A. Tiihonen, F. Oviedo, Z. Liu, J. Thapa, Y. Zhao, N. T. P. Hartono, A. Goyal, T. Heumueller, C. Batali et al., “A data fusion approach to optimize compositional stability of halide perovskites,” Matter 4(4), 1305–1322 (2021). AUTHOR DECLARATIONS 12 Z. Zhang, H. Wang, T. J. Jacobsson, and J. Luo, “Big data driven perovskite solar Conflict of Interest cell stability analysis,” Nat. Commun. 13(1), 7639 (2022). 13 T. J. Jacobsson, A. Hultqvist, A. García-Fernández, A. Anand, A. Al-Ashouri, A. The authors have no conflicts to disclose. Hagfeldt, A. Crovetto, A. Abate, A. G. Ricciardulli, A. Vijayan et al., “An open- access database and analysis tool for perovskite solar cells based on the FAIR data principles,” Nat. Energy 7(1), 107–115 (2022). Author Contributions 14 J. A. Christians, P. A. Miranda Herrera, and P. V. Kamat, “Transformation F.B. and S.R.R. contributed equally to this work. of the excited state and photovoltaic efficiency of CH3 NH3 PbI3 perovskite upon controlled exposure to humidified air,” J. Am. Chem. Soc. 137, 1530 (2015). 15 A. M. A. Leguy, Y. Hu, M. Campoy-Quiles, M. I. Alonso, O. J. Weber, P. Fanny Baumann: Conceptualization (equal); Data curation (equal); Azarhoosh, M. van Schilfgaarde, M. T. Weller, T. Bein, J. Nelson, P. Docampo, Investigation (equal); Writing – original draft (equal); Writing – and P. R. F. Barnes, “Reversible hydration of CH3 NH3 PbI3 in films, single crystals, review & editing (equal). Sonia R. Raga: Conceptualization (equal); and solar cells,” Chem. Mater. 27, 3397 (2015). 16 Data curation (equal); Funding acquisition (equal); Project admin- W. Chi and S. K. Banerjee, “Achieving resistance against moisture and oxygen istration (equal); Supervision (equal); Validation (equal); Writing – for perovskite solar cells with high efficiency and stability,” Chem. Mater. 33(12), original draft (equal); Writing – review & editing (equal). Monica 4269–4303 (2021). 17 Lira-Cantú: Conceptualization (equal); Funding acquisition (equal); Y. Hu, M. F. Aygüler, M. L. Petrus, T. Bein, and P. Docampo, “Impact of rubid- ium and cesium cations on the moisture stability of multiple-cation mixed-halide Project administration (equal); Resources (equal); Supervision perovskites,” ACS Energy Lett. 2, 2212 (2017). (equal); Writing – original draft (equal); Writing – review & editing 18 M. A. A. Kazemi, N. Folastre, P. Raval, M. Sliwa, J. M. V. Nsanzimana, S. (equal). Golonu, A. Demortiere, J. Rousset, O. Lafon, L. Delevoye et al., “Moisture-induced non-equilibrium phase segregation in triple cation mixed halide perovskite mon- DATA AVAILABILITY itored by in situ characterization techniques and solid-state NMR,” Energy Environ. Mater. 6, e12335 (2021). Data sharing is not applicable to this article as no new data were 19 M. A. Akhavan Kazemi, P. Raval, K. Cherednichekno, J. N. Chotard, A. Krishna, created or analyzed in this study. A. Demortiere, G. N. M. Reddy, and F. Sauvage, “Molecular-level insight into APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-29 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape correlation between surface defects and stability of methylammonium lead halide energy in photobrightening and photodegradation of halide perovskite thin films,” perovskite under controlled humidity,” Small Methods 5(2), 2000834 (2021). J. Phys. Chem. Lett. 9(8), 2062–2069 (2018). 20 39 S. Kundu and T. L. Kelly, “Improving the moisture stability of perovskite solar S. D. Stranks, V. M. Burlakov, T. Leijtens, J. M. Ball, A. Goriely, and H. J. Snaith, cells by using PMMA/P3HT based hole-transport layers,” Mater. Chem. Front. “Recombination kinetics in organic-inorganic perovskites: Excitons, free charge, 2(1), 81–89 (2018). and subgap states,” Phys. Rev. Appl. 2(3), 034007 (2014). 21 40 J. Yang, B. D. Siempelkamp, D. Liu, and T. L. Kelly, “Investigation of Y. Tian, M. Peter, E. Unger, M. Abdellah, K. Zheng, T. Pullerits, A. Yart- CH3 NH3 PbI3 degradation rates and mechanisms in controlled humidity environ- sev, V. Sundström, and I. G. Scheblykin, “Mechanistic insights into perovskite ments using in situ techniques,” ACS Nano 9(2), 1955–1963 (2015). photoluminescence enhancement: Light curing with oxygen can boost yield 22 Y. Ying-Guo, Y. Guang-Zhi, F. Shang-Lei, L. Meng, J. Geng-Wu, S. Fei, W. Wen, thousandfold,” Phys. Chem. Chem. Phys. 17(38), 24978–24987 (2015). 41 and G. Xing-Yu, “An in-situ real time study of the perovskite film micro-structural H. Choi, J. C.-R. Ke, S. Skalsky, C. A. Castle, K. Li, K. L. Moore, W. R. Flavell, evolution in a humid environment by using synchrotron based characterization and P. Parkinson, “Visualizing the role of photoinduced ion migration on pho- technique,” Acta Phys. Sin. 66(1), 018401 (2017). toluminescence in halide perovskite grains,” J. Mater. Chem. C 8(22), 7509–7518 23 D. Li, S. A. Bretschneider, V. W. Bergmann, I. M. Hermes, J. Mars, A. Klasen, H. (2020). 42 Lu, W. Tremel, M. Mezger, H.-J. Butt et al., “Humidity-induced grain boundaries M. G. Albrecht and M. Green, “The kinetics of the photolysis of thin films of in MAPbI3 perovskite films,” J. Phys. Chem. C 120(12), 6363–6368 (2016). lead iodide,” J. Phys. Chem. Solids 38, 297 (1977). 24 43 J. Schlipf, L. Bießmann, L. Oesinghaus, E. Berger, E. Metwalli, J. A. Lercher, L. S. Wang, Y. Jiang, E. J. Juarez-Perez, L. K. Ono, and Y. Qi, “Accelerated degrada- Porcar, and P. Müller-Buschbaum, “In situ monitoring the uptake of moisture into tion of methylammonium lead iodide perovskites induced by exposure to iodine hybrid perovskite thin films,” J. Phys. Chem. Lett. 9(8), 2015–2021 (2018). vapour,” Nat. Energy 2(1), 16195 (2016). 25 44 L. Hu, G. Shao, T. Jiang, D. Li, X. Lv, H. Wang, X. Liu, H. Song, J. Tang, and A. Merdasa, A. Kiligaridis, C. Rehermann, M. Abdi-Jalebi, J. Stöber, B. Louis, M. H. Liu, “Investigation of the interaction between perovskite films with moisture Gerhard, S. D. Stranks, E. L. Unger, and I. G. Scheblykin, “Impact of excess lead via in situ electrical resistance measurement,” ACS Appl. Mater. Interfaces 7(45), iodide on the recombination kinetics in metal halide perovskites,” ACS Energy 25113–25120 (2015). Lett. 4(6), 1370–1378 (2019). 26 45 Z. Song, A. Abate, S. C. Watthage, G. K. Liyanage, A. B. Phillips, U. Steiner, M. H. Yuan, E. Debroye, K. Janssen, H. Naiki, C. Steuwe, G. Lu, M. Moris, E. Orgiu, Graetzel, and M. J. Heben, “Perovskite solar cell stability in humid air: Partially H. Uji-i, F. De Schryver et al., “Degradation of methylammonium lead iodide per- reversible phase transitions in the PbI2 -CH3 NH3 I-H2 O system,” Adv. Energy ovskite structures through light and electron beam driven ion migration,” J. Phys. Mater. 6(19), 1600846 (2016). Chem. Lett. 7(3), 561–566 (2016). 27 46 A. N. Urwick, F. Bastianini, G. E. Pérez, and A. Dunbar, “Deciphering perovskite G. M. Paternò, V. Robbiano, L. Santarelli, A. Zampetti, C. Cazzaniga, V. Garcìa decomposition in a humid atmosphere with TOF-GISANS,” Energy Rep. 8, 23–33 Sakai, and F. Cacialli, “Perovskite solar cell resilience to fast neutrons,” Sustainable (2022). Energy Fuels 3(10), 2561–2566 (2019). 28 47 08 November 2023 15:51:28 S. Maniyarasu, B. F. Spencer, H. Mo, A. S. Walton, A. G. Thomas, and W. J. Yang, Q. Hong, Z. Yuan, R. Xu, X. Guo, S. Xiong, X. Liu, S. Braun, Y. Li, J. R. Flavell, “Surface stability of ionic-liquid-passivated mixed-cation perovskite Tang et al., “Unraveling photostability of mixed cation perovskite films in extreme probed with in situ photoelectron spectroscopy,” J. Mater. Chem. A 10(35), environment,” Adv. Opt. Mater. 6(20), 1800262 (2018). 18206–18217 (2022). 48 L. Lu, K.-C. Shen, J. Wang, Z. Su, Y. Li, L. Chen, Y. Luo, F. Song, X. Gao, and 29 Y. Li, X. Xu, C. Wang, B. Ecker, J. Yang, J. Huang, and Y. Gao, “Light-induced J.-X. Tang, “Interaction of the cation and vacancy in hybrid perovskites induced degradation of CH3 NH3 PbI3 hybrid perovskite thin film,” J. Phys. Chem. C 121, by light illumination,” ACS Appl. Mater. Interfaces 12(37), 42369–42377 (2020). 3904 (2017). 49 A. Donakowski, D. W. Miller, N. C. Anderson, A. Ruth, E. M. Sanehira, J. J. 30 A. Senocrate, T. Acartürk, G. Y. Kim, R. Merkle, U. Starke, M. Grätzel, and J. Berry, M. D. Irwin, A. Rockett, and K. X. Steirer, “Improving photostability of Maier, “Interaction of oxygen with halide perovskites,” J. Mater. Chem. A 6(23), cesium-doped formamidinium lead triiodide perovskite,” ACS Energy Lett. 6(2), 10847–10855 (2018). 574–580 (2021). 31 50 N. Aristidou, I. Sanchez-Molina, T. Chotchuangchutchaval, M. Brown, L. Mar- E. T. Hoke, D. J. Slotcavage, E. R. Dohner, A. R. Bowring, H. I. Karunadasa, tinez, T. Rath, and S. A. Haque, “The role of oxygen in the degradation of and M. D. McGehee, “Reversible photo-induced trap formation in mixed-halide methylammonium lead trihalide perovskite photoactive layers,” Angew. Chem., hybrid perovskites for photovoltaics,” Chem. Sci. 6(1), 613–617 (2015). Int. Ed. 54, 8208 (2015). 51 A. J. Barker, A. Sadhanala, F. Deschler, M. Gandini, S. P. Senanayak, P. M. 32 X. Tang, M. Brandl, B. May, I. Levchuk, Y. Hou, M. Richter, H. Chen, S. Pearce, E. Mosconi, A. J. Pearson, Y. Wu, A. R. Srimath Kandada, T. Leijtens, Chen, S. Kahmann, A. Osvet, F. Maier, H.-P. Steinrück, R. Hock, G. J. Matt, F. De Angelis, S. E. Dutton, A. Petrozza, and R. H. Friend, “Defect-assisted pho- and C. J. Brabec, “Photoinduced degradation of methylammonium lead triiodide toinduced halide segregation in mixed-halide perovskite thin films,” ACS Energy perovskite semiconductors,” J. Mater. Chem. A 4, 15896 (2016). Lett. 2, 1416 (2017). 33 52 Y. Nakamura, N. Shibayama, K. Fujiwara, T. Koganezawa, and T. Miyasaka, D. P. McMeekin, G. Sadoughi, W. Rehman, G. E. Eperon, M. Saliba, M. T. “Degradation mechanism of halide perovskite crystals under concurrent light and Hörantner, A. Haghighirad, N. Sakai, L. Korte, B. Rech, M. B. Johnston, L. M. humidity exposure,” ACS Mater. Lett. 4, 2409–2414 (2022). Herz, and H. J. Snaith, “A mixed-cation lead mixed-halide perovskite absorber for 34 K. Ho, M. Wei, E. H. Sargent, and G. C. Walker, “Grain transformation and tandem solar cells,” Science 351, 151 (2016). degradation mechanism of formamidinium and cesium lead iodide perovskite 53 C. M. Sutter-Fella, Q. P. Ngo, N. Cefarin, K. L. Gardner, N. Tamura, C. V. Stan, under humidity and light,” ACS Energy Lett. 6(3), 934–940 (2021). W. S. Drisdell, A. Javey, F. M. Toma, and I. D. Sharp, “Cation-dependent light- 35 P. Pistor, A. Ruiz, A. Cabot, and V. Izquierdo-Roca, “Advanced Raman spec- induced halide demixing in hybrid organic–inorganic perovskites,” Nano Lett. troscopy of methylammonium lead iodide: Development of a non-destructive 18(6), 3473–3480 (2018). characterisation methodology,” Sci. Rep. 6(1), 35973 (2016). 54 J. Liang, C. Chen, X. Hu, Z. Chen, X. Zheng, J. Li, H. Wang, F. Ye, M. Xiao, Z. 36 M. C. Brennan, S. Draguta, P. V. Kamat, and M. Kuno, “Light-induced anion Lu et al., “Suppressing the phase segregation with potassium for highly efficient phase segregation in mixed halide perovskites,” ACS Energy Lett. 3(1), 204–213 and photostable inverted wide-band gap halide perovskite solar cells,” ACS Appl. (2018). Mater. Interfaces 12(43), 48458–48466 (2020). 37 55 D. W. deQuilettes, W. Zhang, V. M. Burlakov, D. J. Graham, T. Leijtens, A. A. J. Knight, A. D. Wright, J. B. Patel, D. P. McMeekin, H. J. Snaith, M. B. John- Osherov, V. Bulović, H. J. Snaith, D. S. Ginger, and S. D. Stranks, “Photo-induced ston, and L. M. Herz, “Electronic traps and phase segregation in lead mixed-halide halide redistribution in organic–inorganic perovskite films,” Nat. Commun. 7, perovskite,” ACS Energy Lett. 4(1), 75–84 (2019). 11683 (2016). 56 N. Aihemaiti, Y. Jiang, Y. Zhu, and S. Peng, “Light-induced phase segregation 38 W.-A. Quitsch, D. W. deQuilettes, O. Pfingsten, A. Schmitz, S. Ognjanovic, S. evolution of all-inorganic mixed halide perovskites,” J. Phys. Chem. Lett. 14(1), Jariwala, S. Koch, M. Winterer, D. S. Ginger, and G. Bacher, “The role of excitation 267–272 (2023). APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-30 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 57 76 W. Mao, C. R. Hall, A. S. R. Chesman, C. Forsyth, Y. B. Cheng, N. W. Duffy, T. S. R. Raga, L. K. Ono, and Y. Qi, “Rapid perovskite formation by CH3 NH2 gas- A. Smith, and U. Bach, “Visualizing phase segregation in mixed-halide perovskite induced intercalation and reaction of PbI2 ,” J. Mater. Chem. A 4(7), 2494–2500 single crystals,” Angew. Chem., Int. Ed. 58(9), 2893–2898 (2019). (2016). 58 77 G. C. Halford, Q. Deng, A. Gomez, T. Green, J. M. Mankoff, and R. A. Belisle, B. Yang, O. Dyck, W. Ming, M.-H. Du, S. Das, C. M. Rouleau, G. Duscher, D. B. “Structural dynamics of metal halide perovskites during photoinduced halide Geohegan, and K. Xiao, “Observation of nanoscale morphological and structural segregation,” ACS Appl. Mater. Interfaces 14(3), 4335–4343 (2022). degradation in perovskite solar cells by in situ TEM,” ACS Appl. Mater. Interfaces 59 D. Kim, J. S. Yun, P. Sharma, D. S. Lee, J. Kim, A. M. Soufiani, S. Huang, M. 8(47), 32333–32340 (2016). A. Green, A. W. Y. Ho-Baillie, and J. Seidel, “Light- and bias-induced structural 78 Y.-H. Seo, J. H. Kim, D.-H. Kim, H.-S. Chung, and S.-I. Na, “In situ TEM variations in metal halide perovskites,” Nat. Commun. 10(1), 444 (2019). observation of the heat-induced degradation of single- and triple-cation planar 60 H. Funk, O. Shargaieva, A. Eljarrat, E. L. Unger, C. T. Koch, and D. Abou- perovskite solar cells,” Nano Energy 77, 105164 (2020). Ras, “In situ TEM monitoring of phase-segregation in inorganic mixed halide 79 M. Long, T. Zhang, M. Liu, Z. Chen, C. Wang, W. Xie, F. Xie, J. Chen, G. Li, perovskite,” J. Phys. Chem. Lett. 11(13), 4945–4950 (2020). and J. Xu, “Abnormal synergetic effect of organic and halide ions on the stability 61 U. B. Cappel, S. Svanström, V. Lanzilotto, F. O. L. Johansson, K. Aitola, B. and optoelectronic properties of a mixed perovskite via in situ characterizations,” Philippe, E. Giangrisostomi, R. Ovsyannikov, T. Leitner, A. Föhlisch, S. Svensson, Adv. Mater. 30(28), 1801562 (2018). N. Mårtensson, G. Boschloo, A. Lindblad, and H. Rensmo, “Partially reversible 80 J. A. Aguiar, S. Wozny, T. G. Holesinger, T. Aoki, M. K. Patel, M. Yang, J. J. photoinduced chemical changes in a mixed-ion perovskite material for solar cells,” Berry, M. Al-Jassim, W. Zhou, and K. Zhu, “In situ investigation of the formation ACS Appl. Mater. Interfaces 9, 34970 (2017). and metastability of formamidinium lead tri-iodide perovskite solar cells,” Energy 62 S. J. Yoon, M. Kuno, and P. V. Kamat, “Shift happens. How halide ion defects Environ. Sci. 9(7), 2372–2382 (2016). influence photoinduced segregation in mixed halide perovskites,” ACS Energy 81 Y.-C. Li, C. Ge, H.-Y. Song, P. Wang, X.-R. Ma, and S.-B. Liu, “Analysis of ultra- Lett. 2, 1507 (2017). fast carrier dynamics and steady-state reflectivity on lattice expansion in metal 63 S. J. Yoon, S. Draguta, J. S. Manser, O. Sharia, W. F. Schneider, M. Kuno, and halide perovskite during continuous illumination,” J. Appl. Phys. 132(1), 013102 P. V. Kamat, “Tracking iodide and bromide ion segregation in mixed halide lead (2022). perovskites during photoirradiation,” ACS Energy Lett. 1(1), 290–296 (2016). 82 B. Chen, J. Song, X. Dai, Y. Liu, P. N. Rudd, X. Hong, and J. Huang, “Synergistic 64 R. E. Beal, D. J. Slotcavage, T. Leijtens, A. R. Bowring, R. A. Belisle, W. H. effect of elevated device temperature and excess charge carriers on the rapid Nguyen, G. F. Burkhard, E. T. Hoke, and M. D. McGehee, “Cesium lead halide light-induced degradation of perovskite solar cells,” Adv. Mater. 31(35), 1902413 perovskites with improved stability for tandem solar cells,” J. Phys. Chem. Lett. 7, (2019). 746 (2016). 83 65 J. Borchert, H. Boht, W. Fränzel, R. Csuk, R. Scheer, and P. Pistor, “Structural W. Rehman, R. L. Milot, G. E. Eperon, C. Wehrenfennig, J. L. Boland, H. J. investigation of co-evaporated methyl ammonium lead halide perovskite films Snaith, M. B. Johnston, and L. M. Herz, “Charge-carrier dynamics and mobil- during growth and thermal decomposition using different PbX2 (X = I, Cl) ities in formamidinium lead mixed-halide perovskites,” Adv. Mater. 27(48), precursors,” J. Mater. Chem. A 3(39), 19842–19849 (2015). 08 November 2023 15:51:28 7938–7944 (2015). 84 66 A. Dualeh, P. Gao, S. I. Seok, M. K. Nazeeruddin, and M. Grätzel, S. Chen, Y. Hou, H. Chen, M. Richter, F. Guo, S. Kahmann, X. Tang, T. Stubhan, “Thermal behavior of methylammonium lead-trihalide perovskite photovoltaic H. Zhang, N. Li et al., “Exploring the limiting open-circuit voltage and the volt- light harvesters,” Chem. Mater. 26, 6160 (2014). age loss mechanism in planar CH3 NH3 PbBr3 perovskite solar cells,” Adv. Energy 85 Mater. 6(18), 1600132 (2016). J.-Y. Ma, J. Ding, H.-J. Yan, D. Wang, and J.-S. Hu, “Temperature-dependent 67 local electrical properties of organic–inorganic halide perovskites: In situ KPFM D. Guo, Z. Andaji Garmaroudi, M. Abdi-Jalebi, S. D. Stranks, and T. J. Savenije, and c-AFM investigation,” ACS Appl. Mater. Interfaces 11(24), 21627–21633 “Reversible removal of intermixed shallow states by light soaking in multication (2019). mixed halide perovskite films,” ACS Energy Lett. 4(10), 2360–2367 (2019). 86 68 E. J. Juarez-Perez, Z. Hawash, S. R. Raga, L. K. Ono, and Y. Qi, “Thermal A. Ruth, M. C. Brennan, S. Draguta, Y. V. Morozov, M. Zhukovskyi, B. Janko, P. Zapol, and M. Kuno, “Vacancy-mediated anion photosegregation kinetics degradation of CH3 NH3 PbI3 perovskite into NH3 and CH3 I gases observed by in mixed halide hybrid perovskites: Coupled kinetic Monte Carlo and optical coupled thermogravimetry–mass spectrometry analysis,” Energy Environ. Sci. measurements,” ACS Energy Lett. 3(10), 2321–2328 (2018). 9(11), 3406–3410 (2016). 87 69 T. Duong, H. K. Mulmudi, Y. Wu, X. Fu, H. Shen, J. Peng, N. Wu, H. T. Nguyen, W. Tan, A. R. Bowring, A. C. Meng, M. D. McGehee, and P. C. McIntyre, D. Macdonald, M. Lockrey et al., “Light and electrically induced phase segregation “Thermal stability of mixed cation metal halide perovskites in air,” ACS Appl. and its impact on the stability of quadruple cation high bandgap perovskite solar Mater. Interfaces 10(6), 5485–5491 (2018). 88 cells,” ACS Appl. Mater. Interfaces 9(32), 26859–26866 (2017). N.-K. Kim, Y. H. Min, S. Noh, E. Cho, G. Jeong, M. Joo, S.-W. Ahn, J. S. 70 Z. Yang, A. Rajagopal, S. B. Jo, C.-C. Chueh, S. Williams, C.-C. Huang, J. K. Lee, S. Kim, K. Ihm et al., “Investigation of thermally induced degradation in Katahara, H. W. Hillhouse, and A. K.-Y. Jen, “Stabilized wide bandgap perovskite CH3 NH3 PbI3 perovskite solar cells using in-situ synchrotron radiation analysis,” solar cells by tin substitution,” Nano Lett. 16(12), 7739–7747 (2016). Sci. Rep. 7(1), 4645 (2017). 89 71 F. Lehmann, A. Franz, D. M. Többens, S. Levcenco, T. Unold, A. Taubert, and E. J. Juarez-Perez, L. K. Ono, and Y. Qi, “Thermal degradation of formami- S. Schorr, “The phase diagram of a mixed halide (Br, I) hybrid perovskite obtained dinium based lead halide perovskites into sym-triazine and hydrogen cyanide by synchrotron X-ray diffraction,” RSC Adv. 9(20), 11151–11159 (2019). observed by coupled thermogravimetry-mass spectrometry analysis,” J. Mater. 72 K. Suchan, J. Just, P. Becker, C. Rehermann, A. Merdasa, R. Mainz, I. G. Chem. A 7(28), 16912–16919 (2019). 90 Scheblykin, and E. L. Unger, “Broad distribution of local I/Br ratio in illumi- X. Wang, J. Gong, X. Shan, M. Zhang, Z. Xu, R. Dai, Z. Wang, S. Wang, X. Fang, nated mixed halide perovskite films revealed by correlative X-ray diffraction and and Z. Zhang, “In situ monitoring of thermal degradation of CH3 NH3 PbI3 films photoluminescence,” arXiv:2101.00645 (2021). by spectroscopic ellipsometry,” J. Phys. Chem. C 123(2), 1362–1369 (2019). 73 91 G. Divitini, S. Cacovich, F. Matteocci, L. Cinà, A. Di Carlo, and C. Ducati, R. Wang, J. Xue, L. Meng, J.-W. Lee, Z. Zhao, P. Sun, L. Cai, T. Huang, Z. Wang, “In situ observation of heat-induced degradation of perovskite solar cells,” Nat. Z.-K. Wang et al., “Caffeine improves the performance and thermal stability of Energy 1, 15012 (2016). perovskite solar cells,” Joule 3(6), 1464–1477 (2019). 74 92 T. W. Kim, N. Shibayama, L. Cojocaru, S. Uchida, T. Kondo, and H. Segawa, K. W. Tan, D. T. Moore, M. Saliba, H. Sai, L. A. Estroff, T. Hanrath, H. J. Snaith, “Real-time in situ observation of microstructural change in organometal halide and U. Wiesner, “Thermally induced structural evolution and performance of perovskite induced by thermal degradation,” Adv. Funct. Mater. 28(42), 1804039 mesoporous block copolymer-directed alumina perovskite solar cells,” ACS Nano (2018). 8(5), 4730–4739 (2014). 75 93 Z. Fan, H. Xiao, Y. Wang, Z. Zhao, Z. Lin, H.-C. Cheng, S.-J. Lee, G. Wang, C. Li, A. Guerrero, S. Huettner, and J. Bisquert, “Unravelling the role of vacan- Z. Feng, W. A. Goddard et al., “Layer-by-layer degradation of methylammonium cies in lead halide perovskite through electrical switching of photoluminescence,” lead tri-iodide perovskite microplates,” Joule 1(3), 548–562 (2017). Nat. Commun. 9(1), 5113 (2018). APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-31 © Author(s) 2023 APL Energy REVIEW scitation.org/journal/ape 94 111 Q. Jeangros, M. Duchamp, J. Werner, M. Kruth, R. E. Dunin-Borkowski, C. Das, M. Wussler, T. Hellmann, T. Mayer, and W. Jaegermann, “In situ XPS B. Niesen, C. Ballif, and A. Hessler-Wyser, “In situ TEM analysis of study of the surface chemistry of MAPI solar cells under operating conditions in organic–inorganic metal-halide perovskite solar cells under electrical bias,” Nano vacuum,” Phys. Chem. Chem. Phys. 20(25), 17180–17187 (2018). 112 Lett. 16(11), 7013–7018 (2016). M. Kot, L. Kegelmann, H. Köbler, M. Vorokhta, C. Escudero, P. Kúš, B. Šmíd, 95 M. Tallarida, S. Albrecht, A. Abate et al., “In situ near-ambient pressure X-ray M.-c. Kim, N. Ahn, D. Cheng, M. Xu, S.-Y. Ham, X. Pan, S. J. Kim, Y. Luo, D. P. Fenning, D. H. S. Tan et al., “Imaging real-time amorphization of hybrid photoelectron spectroscopy reveals the influence of photon flux and water on the perovskite solar cells under electrical biasing,” ACS Energy Lett. 6(10), 3530–3537 stability of halide perovskite,” ChemSusChem 13(21), 5722–5730 (2020). 113 (2021). M. Ahmadi, E. S. Muckley, I. N. Ivanov, M. Lorenz, X. Li, O. Ovchinnikova, 96 M. Ahmadi, L. Collins, K. Higgins, D. Kim, E. Lukosi, and S. V. Kalinin, E. D. Lukosi, J. T. Tisdale, E. Blount, I. I. Kravchenko et al., “Environmental gat- “Spatially resolved carrier dynamics at MAPbBr3 single crystal–electrode ing and galvanic effects in single crystals of organic–inorganic halide perovskites,” interface,” ACS Appl. Mater. Interfaces 11(44), 41551–41560 (2019). ACS Appl. Mater. Interfaces 11(16), 14722–14733 (2019). 114 97 Y. Yuan, T. Li, Q. Wang, J. Xing, A. Gruverman, and J. Huang, “Anomalous S. G. Motti, D. Meggiolaro, A. J. Barker, E. Mosconi, C. A. R. Perini, J. M. Ball, M. Gandini, M. Kim, F. De Angelis, and A. Petrozza, “Controlling competing photovoltaic effect in organic-inorganic hybrid perovskite solar cells,” Sci. Adv. photochemical reactions stabilizes perovskite solar cells,” Nat. Photonics 13(8), 3(3), e1602164 (2017). 98 532–539 (2019). I. E. Gould, C. Xiao, J. B. Patel, and M. D. McGehee, “In-operando charac- 115 S. S. Dipta and A. Uddin, “Stability issues of perovskite solar cells: A critical terization of P-I-N perovskite solar cell under reverse bias,” in 2021 IEEE 48th review,” Energy Technol. 9(11), 2100560 (2021). Photovoltaic Specialists Conference (PVSC), 20–25 June 2021, USA (IEEE, 2021). 116 99 T. A. Chowdhury, M. A. Bin Zafar, M. Sajjad-Ul Islam, M. Shahinuzzaman, J. L. Garrett, E. M. Tennyson, M. Hu, J. Huang, J. N. Munday, and M. S. Leite, M. A. Islam, and M. U. Khandaker, “Stability of perovskite solar cells: Issues and “Real-time nanoscale open-circuit voltage dynamics of perovskite solar cells,” prospects,” RSC Adv. 13(3), 1787–1810 (2023). Nano Lett. 17(4), 2554–2560 (2017). 117 100 R. Wang, M. Mujahid, Y. Duan, Z. K. Wang, J. Xue, and Y. Yang, “A review of E. M. Tennyson, J. M. Howard, B. Roose, J. L. Garrett, J. N. Munday, A. Abate, perovskites solar cell stability,” Adv. Funct. Mater. 29(47), 1808843 (2019). and M. S. Leite, “The effects of incident photon energy on the time-dependent 118 S. Mazumdar, Y. Zhao, and X. Zhang, “Stability of perovskite solar cells: voltage response of lead halide perovskites,” Chem. Mater. 31(21), 8969–8976 Degradation mechanisms and remedies,” Front. Electron. 2, 712785 (2021). (2019). 119 M. Hadadian, J.-H. Smått, and J.-P. Correa-Baena, “The role of carbon-based 101 Y. Liu, N. Borodinov, L. Collins, M. Ahmadi, S. V. Kalinin, O. S. Ovchinnikova, materials in enhancing the stability of perovskite solar cells,” Energy Environ. Sci. and A. V. Ievlev, “Role of decomposition product ions in hysteretic behavior of 13(5), 1377–1407 (2020). metal halide perovskite,” ACS Nano 15(5), 9017–9026 (2021). 120 E. Unger and T. J. Jacobsson, “The perovskite database project: A perspective 102 Z. Ni, H. Jiao, C. Fei, H. Gu, S. Xu, Z. Yu, G. Yang, Y. Deng, Q. Jiang, Y. Liu on collective data sharing,” ACS Energy Lett. 7(3), 1240–1245 (2022). et al., “Evolution of defects during the degradation of metal halide perovskite solar 121 E. Velilla Hernández, J. B. Cano Quintero, J. F. Montoya, I. Mora-Seró, and F. 08 November 2023 15:51:28 cells under reverse bias and illumination,” Nat. Energy 7(1), 65–73 (2022). Jaramillo Isaza, “Outdoor performance of perovskite photovoltaic technology,” in 103 L. Najafi, S. Bellani, L. Gabatel, M. I. Zappia, A. Di Carlo, and F. Bonaccorso, Thin Films Photovoltaics, edited by Z. Beddiaf and S. Chander (IntechOpen, 2021), “Reverse-bias and temperature behaviors of perovskite solar cells at extended Chap. 5. 122 voltage range,” ACS Appl. Energy Mater. 5(2), 1378–1384 (2022). E. Velilla, F. Jaramillo, and I. Mora-Seró, “High-throughput analysis of the ide- 104 ality factor to evaluate the outdoor performance of perovskite solar minimodules,” A. R. Bowring, L. Bertoluzzi, B. C. O’Regan, and M. D. McGehee, “Reverse bias behavior of halide perovskite solar cells,” Adv. Energy Mater. 8(8), 1702365 Nat. Energy 6(1), 54–62 (2021). 123 (2018). M. Jošt, B. Lipovšek, B. Glažar, A. Al-Ashouri, K. Brecl, G. Matič, A. Magome- 105 A. Alberti, I. Deretzis, G. Mannino, E. Smecca, S. Sanzaro, Y. Numata, T. dov, V. Getautis, M. Topič, and S. Albrecht, “Perovskite solar cells go outdoors: Miyasaka, and A. La Magna, “Revealing a discontinuity in the degradation Field testing and temperature effects on energy yield,” Adv. Energy Mater. 10(25), behavior of CH3 NH3 PbI3 during thermal operation,” J. Phys. Chem. C 121(25), 2000454 (2020). 124 13577–13585 (2017). M. Saliba, M. Stolterfoht, C. M. Wolff, D. Neher, and A. Abate, “Measuring 106 A. Alberti, I. Deretzis, G. Mannino, E. Smecca, F. Giannazzo, A. Listorti, S. aging stability of perovskite solar cells,” Joule 2, 1019 (2018). 125 Colella, S. Masi, and A. La Magna, “Nitrogen soaking promotes lattice recovery in W. Song and T. Aernouts, “Novel test scenarios needed to validate outdoor polycrystalline hybrid perovskites,” Adv. Energy Mater. 9(12), 1803450 (2019). stability of perovskite solar cells,” J. Phys.: Energy 2(2), 021003 (2020). 126 107 E. J. Juarez-Perez, L. K. Ono, M. Maeda, Y. Jiang, Z. Hawash, and Y. Qi, Q. Emery, M. Remec, G. Paramasivam, S. Janke, J. Dagar, C. Ulbrich, R. Schlat- mann, B. Stannowski, E. Unger, and M. Khenkin, “Encapsulation and outdoor “Photodecomposition and thermal decomposition in methylammonium halide testing of perovskite solar cells: Comparing industrially relevant process with a lead perovskites and inferred design principles to increase photovoltaic device simplified lab procedure,” ACS Appl. Mater. Interfaces 14(4), 5159–5167 (2022). stability,” J. Mater. Chem. A 6(20), 9604–9612 (2018). 127 108 H. Köbler, M. V. Khenkin, R. Roy, N. Phung, Q. Emery, M. Remec, R. Schlat- F. Ebadi, B. Yang, Y. Kim, R. Mohammadpour, N. Taghavinia, A. Hagfeldt, mann, C. Ulbrich, and A. Abate, “The challenge of designing accelerated indoor and W. Tress, “When photoluminescence, electroluminescence, and open-circuit tests to predict the outdoor lifetime of perovskite solar cells,” PREPRINT (Version voltage diverge – light soaking and halide segregation in perovskite solar cells,” 1) available at Research Square, 17 August (2021). J. Mater. Chem. A 9(24), 13967–13978 (2021). 128 109 M. Babics, M. De Bastiani, E. Ugur, L. Xu, H. Bristow, F. Toniolo, W. Raja, T. D. Siegler, W. A. Dunlap-Shohl, Y. Meng, Y. Yang, W. F. Kau, P. P. A. S. Subbiah, J. Liu, L. V. Torres Merino et al., “One-year outdoor operation of Sunkari, C. E. Tsai, Z. J. Armstrong, Y.-C. Chen, D. A. C. Beck et al., “Water- monolithic perovskite/silicon tandem solar cells,” Cell Rep. Phys. Sci. 4(2), 101280 accelerated photooxidation of CH3 NH3 PbI3 perovskite,” J. Am. Chem. Soc. (2023). 144(12), 5552–5561 (2022). 129 W. Tress, K. Domanski, B. Carlsen, A. Agarwalla, E. A. Alharbi, M. Graet- 110 X. Deng, X. Wen, J. Zheng, T. Young, C. F. J. Lau, J. Kim, M. Green, S. Huang, zel, and A. Hagfeldt, “Performance of perovskite solar cells under simulated and A. Ho-Baillie, “Dynamic study of the light soaking effect on perovskite solar temperature-illumination real-world operating conditions,” Nat. Energy 4(7), cells by in-situ photoluminescence microscopy,” Nano Energy 46, 356–364 (2018). 568–574 (2019). APL Energy 1, 011501 (2023); doi: 10.1063/5.0145199 1, 011501-32 © Author(s) 2023