Photobrightening in Lead Halide Perovskites: Observations, Mechanisms, and Future Potential

There has been a meteoric rise in the commercial potential of lead halide perovskite optoelectronic devices since photovoltaic cells and light‐emitting diodes based on these materials were first demonstrated. One key challenge common to each of these optoelectronic devices is the need to suppress nonradiative recombination, a process that limits the maximum achievable efficiency in photovoltaic cells and light‐emitting diodes. In this Progress Report, recent studies that seek to minimize this loss pathway in perovskites through a photobrightening treatment, whereby the luminescence efficiency is enhanced through a light illumination passivation process are examined. The sensitivity of this effect to various experimental considerations is examined, including atmosphere, photon energy, photon dose, and also the role of perovskite composition and morphology; under certain conditions there can even be photodarkening effects. Consideration of these factors is critical to resolve seemingly conflicting literature reports. Proposed mechanisms are scrutinized, revealing that there is now some consensus but further work is needed to identify the specific defects being passivated and elucidate universal mechanisms. Finally, the prospects for these treatments to minimize halide migration and push the properties of polycrystalline films towards those of their single‐crystal counterparts are discussed.


Introduction
Over the past decade, hybrid perovskites have reenergized interest in high-performance, low-cost devices for energy harvesting and light emission thanks to their remarkably clean semiconductor properties and solution processability. [1] The range of possible compositions is an additional attractive feature of perovskites. Facile composition engineering enables particularly across the entire family of materials to truly enable their potential diverse device applications. Minimizing nonradiative voltage losses will require consideration of nonradiative recombination pathways both within the bulk of the solar cell active layer and at the charge collection layer interfaces. [25,26] The active layers in PSCs and LEDs are generally polycrystalline films, which have a sizable density of subgap trap states (≈10 15 cm −3 ) that act as nonradiative recombination centers. [27][28][29] Perovskite single crystals nominally have superior optoelectronic properties compared to their polycrystalline thin film analogues, with reported trap densities in single crystals of ≈10 10 -10 12 cm −3 . [30,31] Although a proof of concept has been very recently shown by Chen et al., [32] the processing of single crystals into thin film diode structures poses significant technical challenges especially for upscaling, and these devices are still far below the performance of polycrystalline counterparts. There is therefore the tantalizing prospect of exploiting the facile processibility of polycrystalline films in a variety of device structures while attaining optoelectronic properties (cf. low defect densities) of single crystals. To do so, one must investigate suitable defect passivation treatments for polycrystalline films that could be applied at several stages during device fabrication. An ideal treatment would target the interiors of individual grains as well as grain boundaries and interfaces to suppress all nonradiative recombination and overcome any chargecarrier transport limitations. In doing so, they would also inhibit ion migration processes that can lead to device operational instability. [22] Such treatments may allow polycrystalline PSCs and related devices to be commercialized at extremely high performance levels.
One promising route to reduce nonradiative recombination and improve charge transport in perovskite films is a light soaking treatment in the presence of oxygen and/or moisture. [33][34][35] This passivation approach has many advantages, including easy in situ application and potential applicability to a range of perovskite compositions. As reported by our group and others, this process, hereafter referred to as photobrightening, can substantially enhance luminescence yields of polycrystalline thin films, [35] reaching extremely low trap densities with corresponding improvements in PSC performance. However, atmospheric molecules such as oxygen and water (and their derivatives) are also known to degrade perovskite films particularly over extended exposure periods. [33,36,37] Therefore, the question over the universality of the effect with different systems and conditions and long-term stability issues could limit the commercial viability of this treatment. In this Progress Report, we summarize key results in the use of light to boost perovskite optoelectronic quality, as well as dissect the current understanding of the underlying mechanisms. From this, we consider how experimental factors, namely, atmosphere, photon dose, and photon energy, affect perovskites with different compositions and morphologies in order to draw conclusions on the universal applicability of this treatment. The opportunities are discussed alongside the limitations of such an approach, and we highlight how we could mimic the positive effects of oxygen and light using alternative strategies that may not lead to long-term stability issues. This work therefore presents a framework that will

Observations of Photobrightening and Photodarkening Effects
Research into the photobrightening effect has been primarily motivated by the need to understand the fundamental mechanisms that govern perovskite light emission. The wide range of photoluminescence (PL) lifetimes and quantum efficiencies reported by different groups for the same nominal perovskite compositions, e.g., MAPbI 3 , has also exacerbated the need to decouple intrinsic and extrinsic effects. In 2014, Stranks et al. reported comparable kinetics between the increase in PL intensity from a MAPbI 3 film during extended photoexcitation in ambient laboratory air conditions, and the concomitant increase in open-circuit voltage of a MAPbI 3 PSC device during illumination (Figure 1a). [27] Corresponding enhancements have also been determined for PL lifetimes under different atmospheres, [38,39] indicating that the photobrightening effect impacts upon the relative fraction of (non-) radiative decay pathways within the samples. Nevertheless, others have reported photodarkening processes in nominally similar structures and even seemingly comparable experimental conditions. [38,40] For example, Gottesman et al. showed that a MAPbI 3 film (encapsulated and thus protected from the environment) subjected to white light illumination undergoes a reversible decrease in PL intensity over time (Figure 1b), [40] with the recovery period taking place only in dark conditions. Both photobrightening and photodarkening effects have been widely reported in the literature; a summary of the observations reported in a selection of studies we discuss in this report is shown in Table 1, alongside key experimental parameters, highlighting the seemingly conflicting behaviors described. This apparent contradiction will be discussed in subsequent sections by considering a myriad of factors that influence the photobrightening processes, where the interplay between mobile ions, external molecular species, and the perovskite film morphology complicates efforts to universally predict the extent of any PL change and its magnitude. We note here that the PL lifetime, ascertained from some characteristic decay of the PL signal over time after excitation (which is in general nonexponential), depends strongly on the recombination regime and therefore excitation fluence. [49,50] Therefore, care is needed in the extraction and interpretation of this parameter, and any value must be reported with reference to the excitation conditions. In this work, we report where possible PL lifetimes from measurements in a trap-limited recombination regime in which increases or decreases in this parameter qualitatively correspond to increases or decreases in luminescence yields.

Sensitivity to Experimental Conditions
A number of reports have shown that the sign (i.e., photobrightening or photodarkening) and extent of the luminescence change strongly depends on the experimental conditions, which we summarize in Figure 2. Galisteo-López et al. and Tian et al. first showed that oxygen-rich environments are critical for the dramatic increases in the PL intensity of MAPbI 3 films under continuous illumination. [33,34] The sensitivity of the surface to environmental gases was highlighted by Fang et al., [44] who showed that the PL of MAPbBr 3 single crystals drops dramatically over time under illumination in vacuo, but then a PL rise was seen upon subsequent exposure to air. Similar atmospheric-dependent behavior in the PL intensity has been reported in CsPbBr 3 . [51,52] Water molecules have also been shown to contribute to similar photobrightening effects as oxygen, and this can even be synergistic in the presence of both oxygen and water molecules. [35] Anaya et al. quantified the extreme sensitivity of the processes to oxygen: in ultrahigh vacuum (UHV) (10 −8 mbar), the PL decreased under illumination, whereas at pressures as low as 1.3 mbar, the photobrightening effects could already be activated (Figure 2a). [36] These results suggest that the photobrightening effects under continuous illumination may require at least a trace quantity of oxygen and/or other molecules. We note that there is an ongoing competition between photobrightening and photodarkening processes during illumination of such films as atmospheric molecules such as oxygen and water and even light are known to degrade perovskite films when continuously exposed to these stressors. [33,36,37,53] Galisteo-López et al. demonstrated that photobrightening in MAPbI 3 in ambient air was Adv. Energy Mater. 2020, 10, 1903109 Figure 1. Summary of data from two reports on photobrightening and photodarkening. a) Open-circuit voltage (black data) rise following illumination with a white-light LED source with intensity creating an equivalent number of excitations to 1 sun illumination, and PL intensity rise (red data) over time for a film photoexcited with a laser source (532 nm continuous wave (CW) laser, ≈60 mW cm −2 , equivalent to 1 sun). Measurements were performed in ambient laboratory air. Reproduced with permission. [27] Copyright 2014, American Physical Society. b) PL changes over time in a dark-light-dark cycle for an encapsulated MAPbI 3 film, showing a reversible decrease in PL upon light illumination with a white light source with an intensity of <10 mW cm −2 (≈0.1 sun). Reproduced with permission. [40] Copyright 2015, American Chemical Society (ACS).
followed by a subsequent drop of the PL after tens of seconds (depending on the incident light intensity), [33] though this is partly reversible when exposed to dry nitrogen. This rise in PL followed by a drop over longer time periods under illumination has also been reported, albeit at a lower darkening rate, in O 2 and N 2 atmospheres (cf. Figure 2a). [33] The photoexcitation conditions also play an important role in determining the extent of photobrightening. Galisteo-Lopez et al. [33] found that photobrightening and photodarkening depend on the intensity of the excitation, and the rate of both processes increases with intensity. This observation suggests that the photon dose is critical in these processes. deQuilettes et al. [38] demonstrated this by illuminating with different excitation densities and monitoring the PL changes over time ( Figure 2b). The time taken for the PL to rise and reach a stabilized emission output depends on the excitation density, where the time to reach the stabilized emission is longer at lower excitation densities. However, when considering the photon dose in each case (Figure 2b, inset) these curves all overlay, revealing a critical photon dose for stabilized emission in these MAPbI 3 films under these atmospheric conditions to be ≈100 J cm −2 . We note that these measurements were performed with relatively moderate vacuum levels (pressure of ≈10 −3 -10 −4 mbar) and thus the trace oxygen present under such conditions leads to the observed net rise in PL, again highlighting the sensitivity of the effect to oxygen (cf. Figure 2a). We note that this sensitivity to oxygen can also be seen in measurements performed under nominal nitrogen conditions, which will ultimately still contain trace concentrations of oxygen species. Sutter-Fella et al. also showed that the transition from photobrightening to photodarkening is dependent on photon dose. [54] The role of photon energy was studied by Quitsch et al. [46] By light soaking MAPbI 3 thin films for the same duration under ambient laboratory air conditions, they observed a photodarkening effect when photoexciting with photons below a critical wavelength of ≈520 nm and a photobrightening effect above this wavelength ( Figure 2c). [46] They proposed that this energy dependence of the PL involves the photoexcitation of residual PbI 2 , where the threshold of ≈520 nm corresponds to the band edge of PbI 2 , and this can trigger a decomposition process for MAPbI 3 . Merdasa et al. found that light soaking in ambient conditions could further exaggerate this energydependent effect as PbI 2 -rich regions on the surface grow over time under illumination. [55] This process leads to vastly different charge-carrier recombination processes through largely inefficient injection of charge carriers into the perovskite surface via a PbI 2 intermediary when photoexciting at shorter wavelengths.
Another crucial factor is not just how much photon dose arrives, but when this dose arrives and whether there is a relaxation (recovery) time between the doses. Motti et al. recently studied the impact of the relaxation time between illumination on the photobrightening and photodarkening effects in MAPbI 3 and MAPbBr 3 films under high vacuum conditions Adv. Energy Mater. 2020, 10,1903109  (<10 −5 mbar, Figure 2d). They found that with a fixed excitation pulse width but increasing length of time between pulses, they could modulate from a photodarkening (relaxation times between pulses of 0.033 ms) to a photobrightening (relaxation times of 1 ms) effect over early times. This behavior suggests that even in the nominal absence of oxygen a photobrightening effect can be observed provided there is sufficient relaxation time between pulses. This result speaks to the ongoing interplay between mechanisms leading to photobrightening and photo darkening even on short timescales. Nevertheless, we note that longer illumination times (and therefore higher effective photon doses) would be required to explore this interplay further, especially to validate whether there would be a net photo darkening process at longer exposure times. Finally, the rate of photobrightening is strongly dependent on temperature. The phenomenon is slowed down substantially at lower temperatures, requiring an order of magnitude larger photon dose to reach a similar stabilized emission output at 190 K than at 300 K. [38]

Sensitivity to Sample Morphology, Composition, and Passivation
The photobrightening effect is also dependent on the perovskite thin film morphology and composition. deQuilettes et al. [38] investigated how microstructure impacts the PL enhancement in polycrystalline MAPbI 3 films under illumination over time. By correlating scanning electron microscopy (SEM) images and confocal PL microscopy images, they showed that lightsoaking preferentially increases the luminescence from regions with higher trap state densities (dark spots in initial PL images) and also induces a redistribution in local emission intensities (Figure 3a,b). These observations hint that perovskite films of better initial quality exhibit more limited photoinduced changes. Similar observations were made by Brenes et al. [45] using in situ micro-PL measurements under different atmospheric environments. Tian et al. showed that heterogeneity within the perovskite, such as morphological grain size variations, affects the extent to which photobrightening takes place; [43] they found that smaller grains exhibited faster PL enhancement than larger grains. They also demonstrated that different local regions within the same perovskite film have vastly different PL enhancement transients (Figure 3c). [34] Other works have reported spatially dependent PL blinking effects with random on-off fluctuations of nonradiative channels that are particularly exaggerated in nanosized crystals; [56][57][58] this blinking effect is also strongly dependent on the surrounding atmosphere and temperature. [59] The ability to enhance the PL in both MAPbI 3   . Reproduced with permission. [36] Copyright 2018, ACS. b) Normalized integrated PL from a MAPbI 3 film over time under illumination at the stated excitation density with 507-nm pulsed excitation (1 MHz excitation rate). Inset shows the same data plotted as a function of the total photon energy dose. Reproduced under the terms of the CC BY license. [38] Copyright 2019, Springer Nature. c) The PL intensity change in a MAPbI 3 film over time under illumination with either 457-nm (100 mW cm −2 ) or 532-nm (400 mW cm −2 ) laser excitation. Measurements were performed under ambient air conditions. Reproduced with permission. [46] Copyright 2018, ACS. d) Integrated PL intensity, normalized at time t = 0, from a MAPbI 3 film held under vacuum (pressure < 10 −5 mbar) with the stated times held in the dark (δ) between illumination with light pulses of duration 200 ns (≈0.5 µJ cm −2 pulse −1 , 450 nm excitation wavelength). Reproduced with permission. [70] Copyright 2019, Springer Nature.
ionic segregation process that leads to a redshifting of the PL emission to lower energies corresponding to iodide-rich sites. [60] Since carriers can accumulate on these low-bandgap regions, fractions of radiative recombination can be very high, and thus there is a large overall increase in the PL intensity. [60][61][62][63] We note that unintentional fractional deviations in the precursor solution stoichiometry can also impact the extent and sign of photobrightening (or darkening) in perovskite films, a variable which would further contribute to the inconsistencies of reported results. [64] It is reported that the PL intensity increases under continuous illumination in oxygen-rich conditions even in multication perovskites such as (FA,MA,Cs)PbI 3 , but the extent and rate of this increase is in general lower compared to the MAPbI 3 analogues, suggesting there is at least to some degree a suppression of the photobrightening or photodark-ening effects in those compositions. [41,61] The mixed-cation systems also in part retard the halide segregation processes, [47,65,7] though we note that the degree of this inhibition depends on fabrication, crystallinity, and experimental conditions. Generally, continuously illuminating mixed-cation, mixed-halide systems in oxygen-rich conditions leads to an overall passivation of deep traps, and a halide segregation process, with both processes contributing to a net increase in luminescence intensity. [61] In addition, these mixed-cation, mixed-halide systems have local spatial chemical heterogeneity leading to small local variations in the bandgap on the order of tens of milli-electronvolts. This variation leads to shallow states and subsequently local accumulation of one type of photoexcited carrier through a photodoping effect. [66] Interestingly, independent of the Adv. Energy Mater. 2020, 10, 1903109   Figure 3. a) Confocal PL map of a MAPbI 3 film with semitransparent SEM image overlaid (scale bar 2 µm). [38] b) PL intensity of the bright (red), intermediate (green), and dark (blue) regions denoted in a) over time under illumination (470 nm laser, 40 MHz, ≈0.03 µJ cm −2 pulse −1 ). Reproduced under the terms of the CC BY license. [38] Copyright 2016, Springer Nature. c) PL enhancement of different local domains (as highlighted) in a MAPbI 3 film over time under illumination (1500 mW cm −2 , 514 nm CW excitation). Adapted with permission. [34] Copyright 2015, Royal Society of Chemistry. d,e) Confocal PL maps of the same region of a MAPbI 3 film before and after in situ light soaking for 10 min under excitation with a 532 nm laser equivalent to ≈10 suns under humid air (45% relative humidity). Insets to (d) and (e): emission stability of the films in dry air before and after the treatment, respectively, under continuous illumination at 532 nm with intensity equivalent to ≈10 suns. Reproduced under the terms of the CC BY license. [35] Copyright 2017, Cell Press.
atmosphere and changes in PL intensity, the influence of these shallow states can be reduced by light-soaking these systems, likely because the initial well intermixed halide distributions (leading to shallow states) are converted to distinctly segregated iodide and bromide regions. [48] It has also been reported that the light-induced changes can be suppressed through surface passivation of films. Abdi-Jalebi et al. observed that the PL intensity and wavelength in potassium-passivated (FA,MA,Cs)Pb(I/Br) 3 films are more stable under illumination compared to unpassivated films, suggesting that the photoinduced migration processes are substantially inhibited in this composition even in ambient air conditions. [9] This highlights that photoinduced changes are most prevalent in materials of lower starting quality (e.g., unpassivated films), whereas in high-quality materials (e.g., passivated films) this effect will be drastically reduced. The impact of passivation is strikingly demonstrated when considering the PL of MAPbI 3 films passivated through a light treatment in humid air, which homogenizes and increases the local PL intensity distribution when comparing PL maps before ( Figure 3d) and after ( Figure 3e) the treatment. During the illumination treatment, there is a pronounced photobrightening effect (Figure 3d, inset). However, once the sample reaches a stabilized emission level and the continuous illumination and humidity components are removed, further illumination shows that the luminescence is stabilized at enhanced levels with respect to the untreated film and no longer changes over time (Figure 3e, inset).
From our collective analysis above, we conclude that photobrightening is a general property of lead halide perovskites but the extent and rate of the effect depends on the experimental conditions (photon dose, photon energy, relaxation time, environment, temperature) and the sample properties (composition, morphology, and degree of passivation). The apparent contradictions in the literature stem from the high sensitivity of the processes to these factors. We note that further studies on phase-stable CsPbX 3 embodiments are needed to elucidate any mechanistic differences in the inorganic lead halide perovskite family.

Mechanistic Insights
As discussed in the previous sections, the photobrightening effect is sensitive to a myriad of factors, which makes it a very complicated phenomenon to understand. Nevertheless, the sensitivity to the conditions detailed above also provide different levers to adjust the processes, and therefore allow us to extract mechanistic insights to rationalize the observed behavior.
The slow rise and/or drop of the PL intensity during light treatment (minutes) in various environments under typical solar excitation charge densities suggest that electronic charge transport and recombination processes cannot exclusively be responsible for the photobrightening effect. For example, if we simply consider the interplay between trap filling (100 ns timescale) and trap depopulation (100 ms timescale as a conservative slow estimate), the system would reach equilibrium within a few hundreds of microseconds. [27,38,67] Rather, such slow timescales are consistent with ionic (primarily halide) migration, oxygen diffusion, and/or chemical reactions on the surface involving halides and adsorption/desorption of oxygen playing crucial contributing roles. deQuilettes et al. compared confocal PL maps with time-of-flight secondaryion-mass-spectrometry (TOF-SIMS) data on the same scan region. They revealed that the photobrightened region is deficient in iodide with respect to the nonilluminated surrounding region (Figure 4a), and that such a light-induced redistribution of halides depends on the photon dose. [38] This is direct evidence for photoinduced ion migration, and the strong correlations between the extent of this effect and the observed photobrightening support the idea that the mechanisms are intertwined. The hindering of the photobrightening effect in mixed cation systems, in which ion migration is significantly suppressed, [7,47,65] is also consistent with the role of ions in the process; reduced defect densities in these mixed cation systems will also contribute to the suppressed effects, with both phenomena ultimately limiting the supply of reactant species and sites described next. The role of ions is also reflected in the strong temperature dependence of the photobrightening processes, with overall slower processes at lower temperatures consistent with slower ionic movement at low temperature. Activation energies for the photobrightening processes of ≈0.2 eV are observed, which is broadly consistent with the activation energies for halide migration in MAPbI 3 of ≈0.1-0.6 eV. [68,69] Although the mechanism of ion migration in lead halide perovskites is still under debate, a plausible scenario involves the ability of halides to migrate through halide defects in the perovskite structure. [70,60,68,69] Motti et al. proposed that the long-lived carrier traps associated with halide defects govern the photobrightening and photodarkening processes (even in the absence of oxygen) by inducing photochemical reactions of mobile ionic species at those sites. At low charge carrier excitation densities (longer excitation wavelengths), low repetition rates (1000 pulses s −1 ), and low temperatures (≈200 K), they propose that photobrightening can take place in MAPbI 3 through electron trapping at iodide interstitial defects and subsequent Frenkel pair trap annihilation (Figure 4b). [70] This is also consistent with the recent mechanism proposed by Mosconi et al. in which light illumination leads to the annihilation of Frenkel pairs. [42] By contrast, light-induced formation of defects in small volumes promoted by high repetition rates and high excitation densities (shorter wavelengths) has been linked to a bimolecular reaction in which two filled iodine traps could react to form an I 2 molecule. Alternatively, and less probable, sequential electron and hole capture at an iodide interstitial defect pair can induce the formation of I 2 . Either way, I 2 constitutes an electron trap that they propose gives rise to the photodarkening effect (Figure 4b). [70] Other works have proposed mechanisms to understand the photobrightening effect related to atmospheric conditions. [33][34][35]45,53,[71][72][73]36] Illumination in general generates defects such as halide vacancies (e.g., through light-induced ion migration) or atomic lead species. [37,38] It has been proposed that molecular oxygen adsorbs extremely well to perovskite surfaces and defect sites (e.g., the generated vacancies), and then photogenerated electrons can convert the oxygen to superoxide www.advancedsciencenews.com Figure 4c). Thus, continuous exposure to light induces the formation of superoxide species at the perovskite surface, at the expense of bromine, carbon, and nitrogen species (Figure 4d, phase I), a process that is partially reversible upon relaxation in the dark under ultrahigh vacuum conditions (Figure 4d, phase II). [36] The adsorption energy of such superoxide species is extremely high, on the order of chemisorption energies, particularly on neutral and negative halide vacancy sites (≈−3.90 eV). [45] Calculations have shown that the O 2 and O 2 − adsorption reduces the deep trap site density in MAPbI 3 . [73] Godding et al. proposed that the superoxide is a precursor in the formation of peroxide species (e.g., hydrogen peroxide), which in turn oxidizes the atomic lead defects (traps) into benign PbO. [71] They supported their mechanism by observing similar increases in luminescence yields and device performance by employing a direct post-treatment exposure of (FA,Cs)Pb(I/Br) 3 films to hydrogen peroxide (i.e., without light). Such a mechanism could also explain the synergistic effects of oxygen and water, [35,45] where moisture will assist the formation of hydrogen peroxide. In addition, moisture may instigate the partial decomposition of the perovskite at grain boundaries into passivating degradation products such as PbI 2 and PbO, thus converting a defective surface into a benign surface. [74][75][76] Such thin surface layers may also suppress intergrain ion transport, which, along with the passivated defects, leads to the stabilization of the emission after the sample is removed from the light and oxygen/humid atmosphere (cf. Figure 3c and d). The crossover from photobrightening to photodarkening is due to a competitive process in which volatile organic products (i.e., MA and/or FA) are created via peroxide formation, and results in an irreversible darkening. [36,71] The slow timescales involved are due to a combination of the interplay between the photobrightening and photodarkening processes, any slow diffusion processes (such as oxygen or ionic migration), and the specific atmospheric conditions limiting reactants.
A common requirement between the models proposed above is that there is a source of charge carriers (e.g., from photoexcitation or strong oxidizing agents) to instigate photochemical reactions to ameliorate defect states. The reactants in these processes are atmospheric species and/or ionic species (halides, vacancies, or interstitials). Nevertheless, there are still a number of questions remaining, and the proposed models do not yet agree on all aspects. One of the key outstanding questions relates to understanding the intrinsic nature of the defects that are being passivated or exaggerated through lightinduced effects. A recent local structural and photophysics study revealed that the local sites associated with nonradiative decay are directly related to lattice strain. [77] These strains are built-in during the film growth and crystallization stages and may relate to grain boundaries. [78] Other recent work suggests Adv. Energy Mater. 2020, 10,1903109  Reproduced under the terms of the CC BY license. [38] Copyright 2016, Springer Nature. b) Proposed photobrightening and photodarkening mechanisms in MAPbI 3 films. The respective vacancies for each ion are noted by V, and interstitials with a subscript i. Reproduced with permission from Springer Nature (2019). [70] c) Mechanism of reaction steps of oxygen with MAPbI 3 . Reproduced under the terms of the CC BY license. [53] Copyright 2017, Springer Nature. d) X-ray photoemission spectroscopy analysis of a MAPbBr 3 film upon photoexcitation under O 2 exposure (phase I) and during recovery in dark under ultrahigh vacuum conditions (phase II). Left panel shows the evolution of the oxygen (gray circles), water (black circles), superoxide (red circles), and oxide (green circles) signals with respect to Pb, while the right panel shows the evolution of bromine (orange circles), carbon (dark green circles), and nitrogen species (blue circles). Reproduced with permission. [36] Copyright 2018, ACS. that light illumination in mild vacuum conditions can at least in part relieve this strain, resulting in enhanced PL properties and performance of photovoltaic devices. [41] Such findings are consistent with recent reports that strain can impact performance and lead to operational instabilities; [79,80] this is also consistent with the proposition that these nonradiative traps are in turn associated with ionic migration. Such strain formation and relief mechanisms and measurements will need to be reconciled with the atmospheric, photon dose, and photon energy measurements to provide a universal picture for these effects. A complete picture to understand the dynamic interplay between ion migration, trap annihilation, and photoinduced chemical reactions in lead halide perovskites will require an in-depth understanding of the interrelated structural, chemical, and photophysical properties across different length scales (macro down to nanoscale). Only through direct experimental evidence to understand these states will we be able to draw definitive conclusions on the associated photochemical reactions.

How to Maximally Exploit These Effects
Although the underlying mechanisms of the photoinduced processes in perovskites are still under debate, the ability to enhance the properties of lead halide perovskites using such treatments under appropriate conditions is clearly established. We now ask how far we can improve the properties of polycrystalline films using such treatments. To do so, we must ensure the maximum photobrightening effects can be captured to achieve emissive and stable films while suppressing unwanted competing photodarkening or other photochemical processes which will dominate over longer term exposure. Thus, a practical strategy is to optimize the light and atmosphere exposure dose but then isolate the sample from these conditions and subsequently measure the treated film and then complete construction of any device comprising these films. Brenes et al. showed that under optimized conditions of humid air (≈40-50% relative humidity) and photon dose and energy (90 J cm −2 with 532 nm excitation), they could increase the internal PL quantum efficiency values to over 80% from an initial value of ≈1% (Figure 5a), along with a more locally homogeneous emission distribution (cf. Figure 3d and e). [35] They observed this PL enhancement concomitant with an enhanced carrier lifetime to >30 µs under ∼1 sun equivalent excitation conditions (compared to 6 µs before treatment) while maintaining carrier mobilities of 70-90 cm 2 V −1 s −1 ascertained using time-resolved microwave conductivity measurements. These values are comparable to those obtained with the best chemical passivation approaches and single-crystal systems. [54,[81][82][83] Furthermore, the associated surface recombination velocity for holes in the treated film (≈0.4 cm s −1 ) is comparable to the best values reported for any semiconductor in (0.15 cm s −1 in 〈111〉 crystalline silicon [84,85] ), which is remarkable considering that these are solution-processed polycrystalline films. PSCs constructed using the treated films as active layers exhibited higher PCEs (19.2%) compared to control PSCs (17.8%). [35] The enhanced performance is also reflected in the faster rise to stabilized power output, and reaching higher values (Figure 5b), which is in turn consistent with the stabilized PL output (cf. Figure 3d and e) and reduced ionic migration in such treated configurations. [22,86,35] We note that similar results were found by Godding et al. [71] in inverted device structures using films exposed to hydrogen peroxide, which is proposed to generate the same effect as photobrightening. Crucially, the photodarkening processes in such passivated structures are also suppressed. Although this does not necessarily mean that ion migration is Figure 5. a) Enhancement in internal PL quantum efficiency (PLQE) of a MAPbI 3 thin film using a light and atmospheric treatment (excitation with 532 nm CW at ≈150 mW cm −2 , ≈45% relative humidity). [35] b) Stabilized power output PCE values of solar cell devices comprised of films treated with light and atmospheric exposure (red) and untreated films (black). Reproduced under the terms of the CC BY license. [35] Copyright 2017, Cell Press. c) PL spectra of a (FA,MA,Cs)Pb(I 0.6 Br 0.4 ) 3 film over time under 532 nm laser illumination (60 mW cm −2 , ambient laboratory air conditions). The inset is the corresponding external PLQE over time. Reproduced under the terms of the CC BY license. [62] Copyright 2019, The Authors. entirely inhibited, it may mean that any background ion motion may not be detrimental for optoelectronic performance; though further long-term stability testing will be required to validate this assertion. Recently, Andaji-Garmaroudi et al. showed that lightsoaking mixed-cation and mixed-halide films leads to the formation of an extremely thin and well-passivated iodide-rich surface layer on which charges accumulate and recombine with very high radiative efficiencies even at low excitation densities; the PL quantum efficiency of the film increases from 3% to over 45% (Figure 5c). [61] These promising demonstrations show the potential for such passivation strategies to enable new device structures and push perovskite semiconductors toward their limits. Further optimization of these treatments may involve employing the light and atmospheric treatments during film formation to ensure better passivation throughout the bulk of the film. [87] Nevertheless, the presence of oxygen and water derivatives (e.g., superoxide), other light-induced products (e.g., molecular halide species such as I 2 ), or strong oxidizing agents (e.g., hydrogen peroxide) in the structure of treated samples may not provide sufficient long-term stability required for commercial applications. As one of our colleagues eloquently put it, "It would be like taking an operating flamethrower into your home." [88] Therefore, such treatments involving light and/or atmospheric molecules may not be the most suitable solution for eventual scale up. This motivates the search for alternative species that mimic the positive effect of oxygen species and/or light-induced effects in passivating trap states without the detrimental effects. There are widespread reports of chemical passivation treatments employed either through use of additives in the precursor solutions or through post-treatments of bulk films. These include long-chain 2D molecules that incorporate into the perovskite surface structure and sulfur, nitrogen or phosphoruscontaining molecules that interact with surface defects. [89] Use of these molecules to bind to undercoordinated lead may impede the formation of I 2 at these sites and associated photodarkening processes. [70] One key advantage of the light treatment is that charge carriers can be generated in selected regions of the film (dictated, e.g., by the excitation wavelength) and can instigate photochemical reactions in those specific regions. Furthermore, the selective binding of species such as molecules to defect sites may allow one to direct on an atomic scale where the photochemical reactions occur, allowing a more complete amelioration of trap states throughout the bulk of the film than can be achieved with chemical surface treatments alone. One could seek to use other small species that could diffuse through the film and selectively bind to defects in a similar way, which also undergo subsequent photochemical reactions as required. Other strategies may involve removing the adsorbed species once the passivation has occurred, provided the passivation does not require the continued presence of the binding species (e.g., if the molecule simply instigates a transformation of that site). In practice, a combination of all of the above may be required to achieve maximum luminescence yields and film stabilities.

Conclusion
In this Progress Report, we scrutinized the literature on photobrightening and photodarkening processes in lead halide perovskites, where photobrightening refers to the enhancement in luminescence efficiency through continued light illumination. We highlighted the sensitivity of the photobrightening effect to experimental conditions such as atmosphere, photon energy, photon dose, and perovskite composition and morphology, which ultimately leads to an interplay between photobrightening and photodarkening and the seemingly conflicting reports in the literature. We dissected the literature to summarize the common elements of mechanisms, namely, a requirement that there is a source of charge carriers (e.g., from photoexcitation or strong oxidizing agents) to instigate photochemical reactions with atmospheric species and/or ionic species (halides, vacancies, or interstitials) to ameliorate defect states. Although there is growing consensus on many aspects, there are a number of outstanding questions that are under debate, particularly related to the nature of the defects being passivated. Finally, the prospects for these treatments for scaleup to push the properties of polycrystalline films toward those of their single-crystal counterparts scaled-up were discussed. We expect that further optimization of these and similar treatments, along with a deeper understanding of underlying mechanisms, will lead to solar and light-emitting device efficiencies approaching their full potential with minimal ionic migration. This may require in situ characterization techniques in which structural, chemical, and photophysical properties can be simultaneously probed and interrelated to provide universal mechanisms.