The Role of Oxygen in the Degradation of Methylammonium Lead Trihalide Perovskite Photoactive Layers

In this paper we report on the influence of light and oxygen on the stability of CH3NH3PbI3 perovskite based photoactive layers. When exposed to both light and dry air the mp-Al2O3/CH3NH3PbI3 photoactive layers decompose yielding methylamine, PbI2 and I2 as products. We show that this degradation is initiated by the reaction of superoxide (O2 ) with the methylammonium moiety of the perovskite absorber. Fluorescent molecular probe studies indicate that the O2 species is generated by the reaction of photoexcited electrons in the perovskite and molecular oxygen. We show that the yield of O2 generation is significantly reduced when the mp-Al2O3 film is replaced with an mp-TiO2 electron extraction and transport layer. The present findings suggest that replacing the methylammonium component in CH3NH3PbI3 to a species without acid protons could improve tolerance to oxygen and enhance stability.

range of device architectures employing such perovskite absorbers have been reported so far, with efficiencies approaching 20%. [1][2][3][4] In spite of this tremendous progress, a number of key issues must be overcome before wide-spread commercialization is possible, mainly, those associated with toxicity of lead and long-term material stability.
Regarding the latter, UV irradiation, [5] and water, are well known to affect the stability of methylammonium lead triiodide (CH3NH3PbI3, MAPbI3, MAPI). [1,6,7] For example, humidity can degrade the CH3NH3PbI3 absorber and reduce photovoltaic device performance by about 80% in 24 h. [6] Recently, a few mechanistic studies on the action of water on CH3NH3PbI3 have been reported, identifying lead (II) iodide as one of the final products, although different degradation pathways were proposed. [8,9] Moreover, we have recently reported that decomposition of CH3NH3PbI3 can also be triggered by the combination of molecular oxygen and light. [10] More specifically, in our previous work we employed transient absorption spectroscopy to monitor the yield of long-lived charge separation in mesoporous TiO2/CH3NH3PbI3/spiro-OMeTAD and Al2O3/CH3NH3PbI3/spiro-OMeTAD photoactive layers before and after exposure to light and oxygen. After the ageing process, the mesoporous Al2O3 based samples showed a rapid and substantial drop in the yield of long-lived charge separation. We attributed this drop in yield to degradation of the CH3NH3PbI3 absorber resulting from a reaction between oxygen and the photoexcited electrons in CH3NH3PbI3. Moreover, this degradation pathway was apparently hindered in the mesoporous TiO2 based samples; here electron injection from the photoexcited state of CH3NH3PbI3 precludes this parasitic decomposition reaction. In this paper, we build on previous work and report on the mechanism of this photodegradation reaction. Specifically, we show that the combined action of light and molecular oxygen on CH3NH3PbI3 photoactive layers results in the decomposition of the perovskite leading to the formation of PbI2, I2 and 3 methylamine. Importantly, we show that this photodegradation reaction is triggered by the action of a reactive oxygen species (superoxide; O2 -) on the organic (i.e. methylammonium) component of the CH3NH3PbI3 perovskite absorber. The present findings indicate that the identification of alternative cations to methylammonium (CH3NH3 + ) may be of crucial importance to the design of environmentally stable organometal trihalide perovskite photovoltaic devices.
In order to investigate whether superoxide generation contributes to the degradation of MAPI solar cells and leads to their reduced photochemical stability under illumination in a moisture-free atmosphere, a molecular fluorescent probe, hydroethidine (HE) was employed. This compound is known to show a characteristic increase in emission at 610 nm upon exposure to the superoxide radical anion. [11] CH3NH3PbI3 films were fabricated onto two glass substrates coated either with mp-TiO2 or mp-Al2O3. The resulting mp-TiO2/CH3NH3PbI3 and mp-Al2O3/CH3NH3PbI3 films were then immersed into a 0.317 µM solution of the HE probe in dry toluene. The photo-degradation conditions consisted of ageing the samples under continuous illumination from a tungsten lamp (~1.5 mWcm -1 ) and dry air flow (21% oxygen content). In the current experiment, the fluorescence emission of the HE probe was monitored at regular time intervals over the course of 1 hour with the mp-TiO2/CH3NH3PbI3 and mp-Al2O3/CH3NH3PbI3 films exposed to light and dry air. that the process is photo-induced. Similarly, when the dry air stream was replaced with nitrogen and illumination introduced there was no detectable increase in fluorescence intensity (red curve, Figure 1b). In contrast, when the mp-TiO2/CH3NH3PbI3 and mp-Al2O3/CH3NH3PbI3 samples were exposed to both dry air and light, the expected increase of fluorescence of the probe was observed, which is consistent with the presence of superoxide in the solution. However, it is apparent from the data presented in Figure 1b that a higher yield of superoxide generation is observed in the mp-Al2O3/CH3NH3PbI3 film as compared to the analogous mp-TiO2/CH3NH3PbI3 sample.
The next question that arises relates to why the mp-Al2O3/CH3NH3PbI3 film exhibits a greater yield of superoxide generation than the mp-TiO2/CH3NH3PbI3 sample. A key difference between using a mesoporous TiO2 and Al2O3 films is that the TiO2 film can accept an electron from the photoexcited perovskite, due to a favorable energy offset at the heterojunction, [12] as shown in Figure 2. This, in turn, would reduce the number of photoexcited electrons in the perovskite available for electron transfer to oxygen, thereby reducing the yield of superoxide generation, which is consistent with the data presented in Figure 1. We hypothesize that superoxide generation in the mp-TiO2/CH3NH3PbI3 film is most likely due to the presence of a CH3NH3PbI3 overlayer as revealed by SEM studies. Support for this assertion comes from cross-sectional scanning electron microscopy studies which revealed the presence of an approximately 50 nm thick overlayer of perovskite in both mp-Al2O3/CH3NH3PbI3 and mp-TiO2/CH3NH3PbI3 films (see supporting information, Figure S1). As such, the reduced rate of production of superoxide in the TiO2 based samples compared to that of the Al2O3 based films could thus arise as a consequence of the electron transfer process from CH3NH3PbI3 to the TiO2 layer being in direct competition with the reaction of the photo-generated electrons with oxygen. Furthermore, it is pertinent to note that our 5 observation of increased superoxide production in Al2O3 samples is in good agreement with our previous work showing a rapid and profound degradation in the yield of photoinduced charge separation in mp-Al2O3/CH3NH3PbI3/spiro-OMeTAD films upon exposure to light and dry air. [10] We now consider the compositional changes in the CH3NH3PbI3 perovskite film resulting from the ageing process. To elucidate the products of the degradation that occurs as a result of the combination of light and dry air, we immersed the substrates in dry toluene under a constant flow of dry oxygen and illumination, thus excluding moisture. Similar to the HE probe experiments, mp-Al2O3 based films were immersed in dry toluene and then a constant flow of dry oxygen and illumination were used to create the controlled degradation conditions. It was noted that after 72 hours exposure to both light and oxygen the CH3NH3PbI3 perovskite films changed colour from dark brown to yellow and the toluene solution turned light pink, most likely due to the formation of iodine. UV-Vis spectroscopy was used to monitor the evolution of iodine in the solution (Figure 3a). The change in the absorption spectra is consistent with the production of iodine, as the detected absorption spectrum resembles the absorption of iodine. [13] The addition of sodium thiosulfate, which can reduce iodine to iodide, turned the solution colourless, which further supports the identification of the product as iodine.
To further analyze the compositional change in the aged films SEM-EDX measurements were performed. EDX analysis showed the Pb:I ratio to deplete from 1:3 to 1:2 in the fresh and degraded samples respectively, as illustrated in Figure 3b.
6 An X-ray diffraction analysis of the samples (Figure 4) reveals the characteristic peaks of methylammonium lead iodide in the fresh sample. [14,15] These peaks are largely diminished in the degraded sample and distinct additional peaks stemming from PbI2 are detected. The formation of PbI2 was additionally confirmed by Raman spectroscopy, which clearly shows the characteristic peaks of PbI2 that coincide with a PbI2 reference sample (see supporting information, Figure S2).
Having identified two of the degradation products, the next question that arises relates to how the superoxide species causes degradation of the perovskite. One possibility is that the superoxide reacts with the organic component of the perovskite (i.e. methylammonium cation). To test this hypothesis and to better understand the degradation mechanism we investigated whether the methylammonium cations are undergoing any transformation. For this purpose, a model reaction was set up and NMR spectroscopy was employed to monitor the transformation. The model reaction consisted of a sample of MAI dissolved in dry ethanol, to which potassium superoxide was added (KO2). The sample was then transferred to an NMR tube to which some deuterated acetone was added for calibration of the spectrum. The reaction proceeded with a colour change, from colourless to orange. Proton NMR spectra of the sample and of a control containing only methylammonium iodide are shown in Figure 5. Both the signal for the NH3 + and the methyl group protons shift up-field upon addition of potassium superoxide, which is consistent with the deprotonation of the ammonium group to form methylamine. The presence of iodine in the sample was again determined by the addition of sodium thiosulfate, which turned the solution from orange to colourless. The pH of the solution was also estimated with pH-paper to be around 10.
This could be accounted for by the potential formation of KOH due to the potassium 7 salt being used in the model reaction. It is therefore plausible that water can be formed in the oxygen and light assisted degradation of CH3NH3PbI3.
Herein we have shown that the degradation of the mp-Al2O3/CH3NH3PbI3 sample in a moisture free environment can be accounted for by the generation of superoxide through electron transfer from the photoexcited perovskite phase to molecular oxygen.
The present findings suggest that the superoxide acts to break down the perovskite by first deprotonating the methylammonium cation as shown by NMR spectroscopy.
Moreover, the deprotonation of the methylammonium cation ultimately leads to the formation of PbI2, as confirmed by XRD and Raman spectroscopy, and iodine, detected by UV-Vis spectroscopy and confirmed by addition of sodium thiosulphate. Based on these observations we propose a possible reaction scheme for the degradation (see Figure 6). It is pertinent that water is expected to be another by-product of decomposition as discussed above. The resulting water could then participate in further degradation pathways, perhaps similar to the ones reported in the literature, where the water would now act as the active species in the degradation or act to hydrate the perovskite. [8,9] The overall effect after a first action of superoxide on the perovskite will be an increase on the decomposition rate, since, even if the CH3NH3PbI3 is carefully encapsulated in a water-free environment, degradation can be triggered by the presence of oxygen and light. It is therefore crucial to ensure that CH3NH3PbI3 perovskite layers, when incorporated in photovoltaic devices, are isolated from both water and oxygen.
Additionally, the implementation of a mesoporous electron-extracting layer (e.g. mp-TiO2) can facilitate the deactivation of the electron transfer process to oxygen, resulting in reduced superoxide generation. Obtaining more efficient electron extraction should lead to enhanced stability in moisture-free environments. Alternatively material design 8 aimed at replacing the methylammonium cation to a species without acid protons could improve stability and may alleviate the need for oxygen barriers.

Experimental Methods
Materials: All chemicals were purchased from Sigma-Aldrich and used as received, except TiO2 nanoparticles from Dyesol and methylammonium iodide, which was synthesized in the lab. to DMSO. This solution was then spin-coated onto the substrates using a consecutive two-step spin program under a nitrogen atmosphere in a glovebox. The first spinning cycle is performed at 1000 rpm for 10 s followed by 5000 rpm for 20 s, as reported by Jeon et al. [16] During the second phase, the substrate was treated with toluene (~350 mL) drop-casting. The films were then annealed at 100 °C for 10 minutes. The perovskite layers were deposited onto mesoporous Al2O3 and mesoporous TiO2 films prepared as follows:    showing that on ageing the molar ratio of Pb:I in the sample changes from 1:3 to 1:2.
Peaks in the spectra not assigned to Pb or I stem from the glass substrate or the chromium coating of the samples to prevent charging during the SEM-EDX analysis. according to literature data. [14,15] The additional peaks in the degraded sample can be assigned to PbI2 (reference patterns: PDF 01-080-1000 for hexagonal and PDF 01-073-1753 for rhombohedral PbI2).