Machine Learning for Perovskites' Reap-Rest-Recovery Cycle

Abstract

High-performing and low-cost photovoltaics (PV) are critical to the continued adoption of renewable energy sources. While promising, perovskite solar materials show a dynamic optoelectronic response when exposed to H2O, O2, bias, temperature, or light that severely impacts their performance, preventing commercialization. We posit a reap-rest-recovery cycle to avoid permanent material degradation and achieve long-term power conversion efficiency through machine learning (ML). First, the influence of each above-mentioned parameter must be investigated individually and in combination, from the nano- to the macroscale. With sufficient data for ML, provided by a shared-knowledge repository, monitoring frameworks for perovskite solar cells will be developed to maximize long-term operation by using predictive methods to determine the ideal pathways to recovery through rest. With these milestones achieved, we expect perovskite PV to reach the 25 years T80 lifetime requirement. Perovskite photovoltaics are efficient and inexpensive, yet their performance is dynamic. In this Perspective, we examine the effects of H2O, O2, bias, temperature, and illumination on device performance and recovery. First, we discuss pivotal experiments that evaluate perovskites' ability to go through a reap-rest-recovery (3R) cycle, and how machine learning (ML) can help identify the optimum values for each operating parameter. Second, we analyze perovskite dynamics and degradation, emphasizing the research challenges surrounding this 3R cycle. We then outline experiments that could identify the impact of environmental factors on recovery for different perovskite compositions. Finally, we propose an ML paradigm for maximizing long-term performance and predicting device performance recovery, including a shared-knowledge repository. By reframing perovskites' optoelectronic transiency within the context of recovery rather than degradation, we highlight a set of research opportunities and the artificial intelligence solutions needed for the commercial adoption of these promising solar cell materials. Perovskite photovoltaics are efficient and inexpensive, yet their performance is dynamic. In this Perspective, we examine the effects of H2O, O2, bias, temperature, and illumination on device performance and recovery. First, we discuss pivotal experiments that evaluate perovskites' ability to go through a reap-rest-recovery (3R) cycle, and how machine learning (ML) can help identify the optimum values for each operating parameter. Second, we analyze perovskite dynamics and degradation, emphasizing the research challenges surrounding this 3R cycle. We then outline experiments that could identify the impact of environmental factors on recovery for different perovskite compositions. Finally, we propose an ML paradigm for maximizing long-term performance and predicting device performance recovery, including a shared-knowledge repository. By reframing perovskites' optoelectronic transiency within the context of recovery rather than degradation, we highlight a set of research opportunities and the artificial intelligence solutions needed for the commercial adoption of these promising solar cell materials. Hybrid organic-inorganic perovskite (HOIP) photovoltaic (PV) devices are an emerging technology with substantial promise, indicated by a record power conversion efficiency (PCE or η) of 23.3%1Kojima A. Teshima K. Shirai Y. Miyasaka T. Organometal halide perovskites as visible-light sensitizers for photovoltaic cells.J. Am. Chem. Soc. 2009; 131: 6050-6051Crossref PubMed Scopus (15167) Google Scholar and an average increase of ∼2.4% PCE/year.1Kojima A. Teshima K. Shirai Y. Miyasaka T. Organometal halide perovskites as visible-light sensitizers for photovoltaic cells.J. Am. Chem. Soc. 2009; 131: 6050-6051Crossref PubMed Scopus (15167) Google Scholar The general perovskite structure is represented as ABX3, with a monovalent cation placed at the A site, a divalent metal, most often Pb2+, at the B site, and a halide or halide mixture (I−, Cl−, or Br−) occupying the X site. Regarding composition, the A site is predominantly organic, typically formamidinium CH3(NH2)2+ and/or methylammonium CH3NH3+. In the last 2 years, researchers have discovered that the addition of small amounts of Cs, and/or Rb, stabilizes the PV thermal and electronic responses.2Wu Y.L. Yan D. Peng J. Duong T. Wan Y.M. Phang S.P. Shen H.P. Wu N.D. Barugkin C. Fu X. et al.Monolithic perovskite/silicon-homojunction tandem solar cell with over 22% efficiency.Energy Environ. Sci. 2017; 10: 2472-2479Crossref Google Scholar, 3Saliba M. Matsui T. Seo J.Y. Domanski K. Correa-Baena J.P. Nazeeruddin M.K. Zakeeruddin S.M. Tress W. Abate A. Hagfeldt A. et al.Cesium-containing triple cation perovskite solar cells: improved stability, reproducibility and high efficiency.Energy Environ. Sci. 2016; 9: 1989-1997Crossref PubMed Google Scholar This stability enhancement results in more than an order of magnitude increase in the PCE lifetime.3Saliba M. Matsui T. Seo J.Y. Domanski K. Correa-Baena J.P. Nazeeruddin M.K. Zakeeruddin S.M. Tress W. Abate A. Hagfeldt A. et al.Cesium-containing triple cation perovskite solar cells: improved stability, reproducibility and high efficiency.Energy Environ. Sci. 2016; 9: 1989-1997Crossref PubMed Google Scholar, 4Christians J.A. Schulz P. Tinkham J.S. Schloemer T.H. Harvey S.P. de Villers B.J.T. Sellinger A. Berry J.J. Luther J.M. Tailored interfaces of unencapsulated perovskite solar cells for > 1,000 hour operational stability.Nat. Energy. 2018; 3: 68-74Crossref Scopus (628) Google Scholar, 5Tsai H. Asadpour R. Blancon J.C. Stoumpos C.C. Durand O. Strzalka J.W. Chen B. Verduzco R. Ajayan P.M. Tretiak S. et al.Light-induced lattice expansion leads to high-efficiency perovskite solar cells.Science. 2018; 360: 67-70Crossref PubMed Scopus (452) Google Scholar Concerning other material options, Pb-free alternatives are also being pursued by incorporating Sn, Ti, or Sb as the B site metal,6Jokar E. Chien C.-H. Fathi A. Rameez M. Chang Y.-H. Diau E.W.-G. Slow surface passivation and crystal relaxation with additives to improve device performance and durability for tin-based perovskite solar cells.Energy Environ. Sci. 2018; 11: 2353-2362Crossref Google Scholar, 7Chen M. Ju M.-G. Carl A.D. Zong Y. Grimm R.L. Gu J. Zeng X.C. Zhou Y. Padture N.P. Cesium titanium(IV) bromide thin films based stable lead-free perovskite solar cells.Joule. 2018; 2: 558-570Abstract Full Text Full Text PDF Scopus (315) Google Scholar, 8Correa-Baena J.-P. Nienhaus L. Kurchin R.C. Shin S.S. Wieghold S. Putri Hartono N.T. Layurova M. Klein N.D. Poindexter J.R. Polizzotti A. et al.A-site cation in inorganic A3Sb2I9 perovskite influences structural dimensionality, exciton binding energy, and solar cell performance.Chem. Mater. 2018; 30: 3734-3742Crossref Scopus (99) Google Scholar in order to allay toxicological concerns.9Abate A. Perovskite solar cells go lead free.Joule. 2017; 1: 659-664Abstract Full Text Full Text PDF Scopus (230) Google Scholar Despite the above-mentioned meteoric rise in performance, HOIPs present dynamic electrical10Garrett J.L. Tennyson E.M. Hu M. Huang J. Munday J.N. Leite M.S. Real-time nanoscale open-circuit voltage dynamics of perovskite solar cells.Nano Lett. 2017; 17: 2554-2560Crossref PubMed Scopus (90) Google Scholar and optical11Howard J.M. Tennyson E.M. Barik S. Szostak R. Waks E. Toney M.F. Nogueira A.F. Neves B.R.A. Leite M.S. Humidity-induced photoluminescence hysteresis in variable Cs/Br ratio hybrid perovskites.J. Phys. Chem. Lett. 2018; 9: 3463-3469Crossref PubMed Scopus (40) Google Scholar responses and, often, critical instabilities under the intrinsic and extrinsic working conditions shown in Figure 1A. The effects of extrinsic parameters (the presence of H2O and O2) can be potentially mitigated through suitable encapsulation and fabrication strategies. Conversely, the intrinsic parameters are unavoidable during device operation, and defined here as bias, temperature, and light. Therefore, identifying, understanding, and controlling the influence of each one of these factors (as well as their combined effects) toward the stability of HOIPs from the macro- to the nanoscale will continue to be a major thrust in the research community. For instance, there are 31 possible combinations between the five parameters (for each perovskite chemical composition), without considering the order of exposure and the range of values for each. In our opinion, the use of machine learning (ML) is essential to track and predict the influence of each intrinsic and extrinsic parameter on the performance of perovskite solar cells. Thus, the realization of stable perovskite PVs that can deliver reliable power will certainly benefit from the implementation of artificial intelligence (AI) computational methods, as we discuss later. In this Perspective, we discuss the pressing need for additional research into perovskites to identify and control the reap-rest-recovery (3R) cycle through ML, in both established and emerging material combinations (e.g., Pb-free options) that will lead to reliable PV devices (see Figure 1B). Here, we define recovery in a solar cell device as the ability to restore its PCE after a given amount of time spent under resting conditions, e.g., in the absence of light and bias. A perovskite device initially performs under standard operating conditions, defined here as the reap part of the cycle, where the solar energy is harvested and produces the cell's output power. However, the electrical efficiency of these devices usually deteriorates as a function of time and, therefore, it needs to enter the second phase of the cycle: rest, to avoid permanent material degradation. Given a sufficient rest period under appropriate conditions, the solar cell will have completed the cycle, as it optimally recovers its initial power output. Then, the reap phase begins again. Lastly, we address in detail the powerful role of ML methods for uncovering the ideal 3R operating parameters for HOIP PV over 100 s and eventually 1,000 s of performance cycles. Through this contribution, we provide a framework for an ML approach for obtaining reliable HOIP PV solar cells, which could be expanded to commercial modules. Additional research, both fundamental and applied, is required to fully understand and control the dynamics throughout the 3R cycle in state-of-the-art perovskite PV. Degradation in this class of materials has been viewed as a challenge to surmount, with modest attention to the existence and enhancement of performance recovery. HOIP devices of various absorber-layer compositions now have the ability to perform for >1,000 hr,4Christians J.A. Schulz P. Tinkham J.S. Schloemer T.H. Harvey S.P. de Villers B.J.T. Sellinger A. Berry J.J. Luther J.M. Tailored interfaces of unencapsulated perovskite solar cells for > 1,000 hour operational stability.Nat. Energy. 2018; 3: 68-74Crossref Scopus (628) Google Scholar, 5Tsai H. Asadpour R. Blancon J.C. Stoumpos C.C. Durand O. Strzalka J.W. Chen B. Verduzco R. Ajayan P.M. Tretiak S. et al.Light-induced lattice expansion leads to high-efficiency perovskite solar cells.Science. 2018; 360: 67-70Crossref PubMed Scopus (452) Google Scholar, 6Jokar E. Chien C.-H. Fathi A. Rameez M. Chang Y.-H. Diau E.W.-G. Slow surface passivation and crystal relaxation with additives to improve device performance and durability for tin-based perovskite solar cells.Energy Environ. Sci. 2018; 11: 2353-2362Crossref Google Scholar without dramatic performance losses (T80 > 2,000 hr).12Domanski K. Alharbi E.A. Hagfeldt A. Gratzel M. Tress W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells.Nat. Energy. 2018; 3: 61-67Crossref Scopus (455) Google Scholar Importantly, resting the device without illumination can restore the power output of some devices to >95% initial values (assuming an inert environment).13Domanski K. Roose B. Matsui T. Saliba M. Turren-Cruz S.H. Correa-Baena J.P. Carmona C.R. Richardson G. Foster J.M. De Angelis F. et al.Migration of cations induces reversible performance losses over day/night cycling in perovskite solar cells.Energy Environ. Sci. 2017; 10: 604-613Crossref Google Scholar Regardless of these initial experiments, the amount of work focusing on degradation mechanisms far outweighs the quantity addressing recovery pathways in perovskites of different chemical compositions. Device performance in HOIP is often path dependent with respect to ambient conditions,11Howard J.M. Tennyson E.M. Barik S. Szostak R. Waks E. Toney M.F. Nogueira A.F. Neves B.R.A. Leite M.S. Humidity-induced photoluminescence hysteresis in variable Cs/Br ratio hybrid perovskites.J. Phys. Chem. Lett. 2018; 9: 3463-3469Crossref PubMed Scopus (40) Google Scholar and degradation studies rarely optimize rest and/or recovery steps. In this section, we outline how micro- and macroscopic methods have been used to tackle perovskite dynamics under distinct environmental factors, emphasizing rest and recovery when appropriate. This discussion is followed by suggestions for future experiments that can provide a robust description of the transient optoelectronic behavior across the entire 3R cycle. Due to the extensive number of perovskites suitable for PV (>9,000),14Takahashi K. Takahashi L. Miyazato I. Tanaka Y. Searching for hidden perovskite materials for photovoltaic systems by combining data science and first principle calculations.ACS Photon. 2018; 5: 771-775Crossref Scopus (54) Google Scholar and how each chemical composition has distinct stability limits (i.e., performance response when exposed to the intrinsic and extrinsic parameters displayed in Figure 1A), the implementation of supervised and unsupervised ML routines for the experiments highlighted in this section will enable timely feedback about the conditions for optimizing both rest and recovery. As expected, an extensive variety of macroscopic measurements addressing the primary factors affecting perovskite dynamics, H2O, O2, bias, temperature, and illumination, have been performed.15Berhe T.A. Su W.-N. Chen C.-H. Pan C.-J. Cheng J.-H. Chen H.-M. Tsai M.-C. Chen L.-Y. Dubale A.A. Hwang B.-J. Organometal halide perovskite solar cells: degradation and stability.Energy Environ. Sci. 2016; 9: 323-356Crossref Google Scholar, 16Deretzis I. Smecca E. Mannino G. La Magna A. Miyasaka T. Alberti A. Stability and degradation in hybrid perovskites: is the glass half-empty or half-full?.J. Phys. Chem. Lett. 2018; 9: 3000-3007Crossref PubMed Scopus (81) Google Scholar In Figure 2, we highlight a subset wherein the 3R cycle has been partially addressed. For example, device efficiency half-life depends substantially on the surrounding ambient, with samples aged in N2 lasting more than 60× longer than those aged in air containing 100% relative humidity (rH) (see Figure 2A––representing the reap phase). This implies that HOIP solar cells performing in an inert or encapsulated environment can perform longer before needing to rest, as anticipated. Because device performance often decreases as a function of time due to material degradation, the devices must rest for an amount of time that depends on the perovskites' chemical composition, and the environment (including the five parameters displayed in Figure 1A). A rest process for an HOIP solar cell based on a CH3NH3PbI3-xClx absorber is shown in Figure 2B. Here, the residual photovoltage in the dark results from the migration of ionic species to the electron transport layer interface, where they act as recombination sites.17Hu J.G. Gottesman R. Gouda L. Kama A. Priel M. Tirosh S. Bisquert J. Zaban A. Photovoltage behavior in perovskite solar cells under light-soaking showing photoinduced interfacial changes.ACS Energy Lett. 2017; 2: 950-956Crossref Scopus (71) Google Scholar The rest phase time for this voltage condition strongly depends on the electron transport layer (TiO2 or Al2O3) and injection levels (named low-I0 and high-I0). Concerning recovery, this phase heavily depends on the rest conditions. When properly rested, all figures-of-merit of the device can be restored to >70% of their initial values (see Figure 2C for a recovery example).18Khenkin M.V. Annop K.M. Visoly-Fisher I. Kolusheva S. Galagan Y. Di Giacomo F. Vukovic O. Patil B.R. Sherafatipour G. Turkovic V. et al.Dynamics of photoinduced degradation of perovskite photovoltaics: from reversible to irreversible processes.ACS Appl. Energy Mater. 2018; 1: 799-806Crossref Scopus (69) Google Scholar This work shows that the required duration of the rest phase depends on how much performance is lost during the actual solar cell operation. For an HOIP device that decays to 80% (aged to its T80 lifetime) of its initial PCE (in blue in the left graph), only 3 hr of rest returns all figures-of-merit to >90% of their starting values. Contrastingly, the same device at 50% of initial power output (see right graph, corresponding to T50) needs more than 30 hr to recover to ∼80%.18Khenkin M.V. Annop K.M. Visoly-Fisher I. Kolusheva S. Galagan Y. Di Giacomo F. Vukovic O. Patil B.R. Sherafatipour G. Turkovic V. et al.Dynamics of photoinduced degradation of perovskite photovoltaics: from reversible to irreversible processes.ACS Appl. Energy Mater. 2018; 1: 799-806Crossref Scopus (69) Google Scholar Researchers have also identified HOIP cells with contrary fatigue behavior that recover their performance while under illumination, instead of under dark contitions.19Huang F. Jiang L. Pascoe A.R. Yan Y. Bach U. Spiccia L. Cheng Y.-B. Fatigue behavior of planar CH3NH3PbI3 perovskite solar cells revealed by light on/off diurnal cycling.Nano Energy. 2016; 27: 509-514Crossref Scopus (65) Google Scholar Moreover, the restoration behavior can strongly depend on the environmental conditions during aging, where both dark-recovery and light-recovery can occur.20Khenkin M.V. Anoop K.M. Visoly-Fisher I. Galagan Y. Di Giacomo F. Patil B.R. Sherafatipour G. Turkovic V. Rubahn H.G. Madsen M. et al.Reconsidering figures of merit for performance and stability of perovskite photovoltaics.Energy Environ. Sci. 2018; 11: 739-743Crossref Google Scholar These results emphasize the need to comprehensively explore recovery under the five parameters displayed in Figure 1A (both in isolation and when combined) for >100 hr. Because the time and conditions for an effective rest phase strongly depends on the type and the “usage” of the cells, supervised ML is ideal to help deciding the precise values of the extrinsic and intrinsic parameters, as will be discussed in the next section. To date, the impact of the extrinsic environmental parameters (H2O and O2) on the prototypical methylammonium lead triiodide (MAPbI3) composition is relatively well understood compared with other perovskite chemical compositions. Under illumination, O2 initially passivates this perovskite's defects, but also promotes deterioration of the device's optoelectronic properties.21Aristidou N. Eames C. Sanchez-Molina I. Bu X. Kosco J. Islam M.S. Haque S.A. Fast oxygen diffusion and iodide defects mediate oxygen-induced degradation of perovskite solar cells.Nat. Commun. 2017; 8: 15218Crossref PubMed Scopus (699) Google Scholar The deleterious effect of O2, with and without light, is far greater than that of N2.22Bryant D. Aristidou N. Pont S. Sanchez-Molina I. Chotchunangatchaval T. Wheeler S. Durrant J.R. Haque S.A. Light and oxygen induced degradation limits the operational stability of methylammonium lead triiodide perovskite solar cells.Energy Environ. Sci. 2016; 9: 1655-1660Crossref Google Scholar Macroscopically, the influence of rH has been determined across a range of cation and halide compositions in HOIPs, capturing the rate at which the figures-of-merit diminish.4Christians J.A. Schulz P. Tinkham J.S. Schloemer T.H. Harvey S.P. de Villers B.J.T. Sellinger A. Berry J.J. Luther J.M. Tailored interfaces of unencapsulated perovskite solar cells for > 1,000 hour operational stability.Nat. Energy. 2018; 3: 68-74Crossref Scopus (628) Google Scholar, 12Domanski K. Alharbi E.A. Hagfeldt A. Gratzel M. Tress W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells.Nat. Energy. 2018; 3: 61-67Crossref Scopus (455) Google Scholar, 23Huang W. Manser J.S. Kamat P.V. Ptasinska S. Evolution of chemical composition, morphology, and photovoltaic efficiency of CH3NH3PbI3 perovskite under ambient conditions.Chem. Mater. 2016; 28: 303-311Crossref Scopus (158) Google Scholar, 24Christians J.A. Miranda Herrera P.A. Kamat P.V. Transformation of the excited state and photovoltaic efficiency of CH3NH3PbI3 perovskite upon controlled exposure to humidified air.J. Am. Chem. Soc. 2015; 137: 1530-1538Crossref PubMed Scopus (1023) Google Scholar In addition, with in situ X-ray diffraction (XRD) analyses, the timescales for emergence of hydrate phases were captured.25Leguy A.M.A. Hu Y. Campoy-Quiles M. Alonso M.I. Weber O.J. Azarhoosh P. van Schilfgaarde M. Weller M.T. Bein T. Nelson J. et al.Reversible hydration of CH3NH3PbI3 in films, single crystals, and solar cells.Chem. Mater. 2015; 27: 3397-3407Crossref Scopus (955) Google Scholar In situ photoluminescence (PL) spectroscopy has also revealed the time-dependence of a perovskite's bandgap on exposure to humidified N2.26Hall G.N. Stuckelberger M. Nietzold T. Hartman J. Park J.-S. Werner J. Niesen B. Cummings M.L. Rose V. Ballif C. et al.The role of water in the reversible optoelectronic degradation of hybrid perovskites at low pressure.J. Phys. Chem. C. 2017; 121: 25659-25665Crossref Scopus (17) Google Scholar Research investigating bias and light (intrinsic parameters) on MAPbI3 found an approximate doubling in both short-circuit current density (Jsc) and PCE within 2 min of operation.27Snaith H.J. Abate A. Ball J.M. Eperon G.E. Leijtens T. Noel N.K. Stranks S.D. Wang J.T. Wojciechowski K. Zhang W. Anomalous hysteresis in perovskite solar cells.J. Phys. Chem. Lett. 2014; 5: 1511-1515Crossref PubMed Scopus (2026) Google Scholar Long-term measurements have shown that holding triple-cation perovskite devices at Voc, Jsc, and maximum power operation point for <100 hr yields PCE retention of ∼55%, >60%, and 75%, respectively.12Domanski K. Alharbi E.A. Hagfeldt A. Gratzel M. Tress W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells.Nat. Energy. 2018; 3: 61-67Crossref Scopus (455) Google Scholar Using temperature as the variable parameter (also an intrinsic factor), researchers have identified the negative effects of temperature extremes (−10°C to 65°C) during 500 hr of logging12Domanski K. Alharbi E.A. Hagfeldt A. Gratzel M. Tress W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells.Nat. Energy. 2018; 3: 61-67Crossref Scopus (455) Google Scholar and the reversibility of MAPbI3 PCE throughout thermal cycles of different bounds.28Divitini G. Cacovich S. Matteocci F. Cina L. Di Carlo A. Ducati C. In situ observation of heat-induced degradation of perovskite solar cells.Nat. Energy. 2016; 1: 15012Crossref Google Scholar PL spectroscopy has been used extensively to evaluate the dependence of optical stability on both cation and halide composition29Hoke E.T. Slotcavage D.J. Dohner E.R. Bowring A.R. Karunadasa H.I. McGehee M.D. Reversible photo-induced trap formation in mixed-halide hybrid perovskites for photovoltaics.Chem. Sci. 2015; 6: 613-617Crossref PubMed Google Scholar, 30Braly I.L. Stoddard R.J. Rajagopal A. Uhl A.R. Katahara J.K. Jen A.K.Y. Hillhouse H.W. Current-induced phase segregation in mixed halide hybrid perovskites and its impact on two-terminal tandem solar cell design.ACS Energy Lett. 2017; 2: 1841-1847Crossref Scopus (141) Google Scholar and the influence of wavelength on dynamic photodegradation and photobrightening.31Quitsch W.A. deQuilettes D.W. Pfingsten O. Schmitz A. Ognjanovic S. Jariwala S. Koch S. Winterer M. Ginger D.S. Bacher G. The role of excitation energy in photobrightening and photodegradation of halide perovskite thin films.J. Phys. Chem. Lett. 2018; 9: 2062-2069Crossref PubMed Scopus (65) Google Scholar In addition, cycling light excitation density under different ambient environments (vacuum, N2, and air) influences the extent of luminescence recovery, revealing that changes to the PL quantum yield are path and composition dependent.11Howard J.M. Tennyson E.M. Barik S. Szostak R. Waks E. Toney M.F. Nogueira A.F. Neves B.R.A. Leite M.S. Humidity-induced photoluminescence hysteresis in variable Cs/Br ratio hybrid perovskites.J. Phys. Chem. Lett. 2018; 9: 3463-3469Crossref PubMed Scopus (40) Google Scholar, 32Motti S.G. Gandini M. Barker A.J. Ball J.M. Kandada A.R.S. Petrozza A. Photoinduced emissive trap states in lead halide perovskite semiconductors.ACS Energy Lett. 2016; 1: 726-730Crossref Scopus (127) Google Scholar In particular, UV light has been shown to cause degradation of perovskite PV output over the course of 1,000 hr. However, these losses can be ∼40% recoverable over 10 cycles by alternating between UV and AM1.5G illumination.33Lee S.W. Kim S. Bae S. Cho K. Chung T. Mundt L.E. Lee S. Park S. Park H. Schubert M.C. et al.UV degradation and recovery of perovskite solar cells.Sci. Rep. 2016; 6: 38150Crossref PubMed Scopus (220) Google Scholar Note that while we define UV light as an intrinsic parameter, we recognize that it could be considered extrinsic given the option of using UV-blocking encapsulant layers.12Domanski K. Alharbi E.A. Hagfeldt A. Gratzel M. Tress W. Systematic investigation of the impact of operation conditions on the degradation behaviour of perovskite solar cells.Nat. Energy. 2018; 3: 61-67Crossref Scopus (455) Google Scholar The substantial variation in performance as a function of illumination conditions urges the realization of supervised and unsupervised ML on perovskite thin films (prior to full device development), where PL measurements are sufficient to determine the conditions for radiative recombination recovery.11Howard J.M. Tennyson E.M. Barik S. Szostak R. Waks E. Toney M.F. Nogueira A.F. Neves B.R.A. Leite M.S. Humidity-induced photoluminescence hysteresis in variable Cs/Br ratio hybrid perovskites.J. Phys. Chem. Lett. 2018; 9: 3463-3469Crossref PubMed Scopus (40) Google Scholar Because most perovskites present inhomogeneities at the nano- and microscale, microscopic techniques must be further developed to resolve the relationship between composition, morphology, optical response, and electrical behavior at the intragrain and intergrain length scales.34Tennyson E.M. Gong C. Leite M.S. Imaging energy harvesting and storage systems at the nanoscale.ACS Energy Lett. 2017; 2: 2761-2777Crossref Scopus (33) Google Scholar, 35Tennyson E.M. Garrett J.L. Frantz J.A. Myers J.D. Bekele R.Y. Sanghera J.S. Munday J.N. Leite M.S. Nanoimaging of open-circuit voltage in photovoltaic devices.Adv. Energy Mater. 2015; 5: 1501142Crossref Scopus (65) Google Scholar, 36Tennyson E.M. Howard J.M. Leite M.S. Mesoscale functional imaging of materials for photovoltaics.ACS Energy Lett. 2017; 2: 1825-1834Crossref Scopus (26) Google Scholar, 37Leite M.S. Abashin M. Lezec H.J. Gianfrancesco A. Talin A.A. Zhitenev N.B. Nanoscale imaging of photocurrent and efficiency in CdTe solar cells.ACS Nano. 2014; 8: 11883-11890Crossref PubMed Scopus (51) Google Scholar, 38Garrett J.L. Leite M.S. Munday J.N. Multiscale functional imaging of interfaces through atomic force microscopy using harmonic mixing.ACS Appl. Mater. Interfaces. 2018; 10: 28850-28859Crossref PubMed Scopus (12) Google Scholar Figure 3 displays examples of how microscopic methods have been implemented to help elucidate the dynamic response of this emerging material. Through environmentally controlled micro-PL, the effect of ambient gas and vacuum was identified, showing that the presence of O2 can lead to an order of magnitude increase in radiative recombination; however, as shown in Figure 3A, not all grains behave identically and the phenomenon is facet dependent.39Tian Y. Peter M. Unger E. Abdellah M. Zheng K. Pullerits T. Yartsev A. Sundstrom V. Scheblykin I.G. Mechanistic insights into perovskite photoluminescence enhancement: light curing with oxygen can boost yield thousandfold.Phys. Chem. Chem. Phys. 2015; 17: 24978-24987Crossref PubMed Google Scholar Using wide-field PL imaging, the role of an electric field (bias) on ion migration has been captured in real time.40Li C. Guerrero A. Zhong Y. Graser A. Luna C.A.M. Kohler J. Bisquert J. Hildner R. Huettner S. Real-time observation of iodide ion migration in methylammonium lead halide perovskites.Small. 2017; 13: 1701711Crossref Scopus (118) Google Scholar The real-time light-induced dynamics at the nanoscale are accessible through Kelvin-probe force microscopy, identifying intragrain voltage variances of ∼300 mV that decay over 128 s after returning to dark conditions (Figure 3B).10Garrett J.L. Tennyson E.M. Hu M. Huang J. Munday J.N. Leite M.S. Real-time nanoscale open-circuit voltage dynamics of perovskite solar cells.Nano Lett. 2017; 17: 2554-2560Crossref PubMed Scopus (90) Google Scholar Photoconductive atomic force microscopy (pc-AFM) has revealed intragrain variations in MAPbI3 device figures-of-merit that can be correlated with surface microstructure.41Kutes Y. Zhou Y. Bosse J.L. Steffes J. Padture N.P. Huey B.D. Mapping the photoresponse of CH3NH3PbI3 hybrid perovskite thin films at the nanoscale.Nano Lett. 2016; 16: 3434-3441Crossref PubMed Scopus (109) Google Scholar In addition, pc-AFM has been used to image the photoinactive surface regions in temperature-cycled MAPbI3 solar cells.42Conings B. Drijkoningen J. Gauquelin N. Babayigit A. D'Hae