Inﬂuence of Oxygen Ion Migration from Substrates on Photochemical Degradation of CH 3 NH 3 PbI 3 Hybrid Perovskite

: Measurements of XPS survey, core levels (N 1s, O 1s, Pb 4f, I 3d), and valence band (VB) spectra of CH 3 NH 3 PbI 3 (MAPbI 3 ) hybrid perovskite prepared on different substrates (glass, indium tin oxide (ITO), and TiO 2 ) aged under different light-soaking conditions at room temperature are presented. The results reveal that the photochemical stability of MAPbI 3 depends on the type of substrate and gradually decreases when glass is replaced by ITO and TiO 2 . Also, the degradation upon exposure to visible light is accompanied by the formation of MAI, PbI 2 , and Pb 0 products as shown by XPS core levels spectra. According to XPS O 1s and VB spectra measurements, this degradation process is superimposed on the partial oxidation of lead atoms in ITO/MAPbI 3 and TiO 2 /MAPbI 3 , for which Pb–O bonds are formed due to the diffusion of the oxygen ions from the substrates. This unexpected interaction leads to additional photochemical degradation.


Introduction
The possibility of using hybrid perovskites in photovoltaics was demonstrated 10 years ago [1]. Since then, these materials have become the subject of numerous studies and still attract enormous interest due to their excellent optoelectronic properties, low manufacturing costs, tunable bandgaps, and high efficiency [2][3][4][5]. These materials can be solutionprocessed at low temperature and vapor-deposited, realistically holding the promise to reach comparable efficiency as conventional thin-film photovoltaic technologies had. Besides, they can be combined with organic electron or hole transport materials to realize flexible photovoltaic devices. However, the instability of these materials to light, moisture, and temperature has aroused a serious concern that needs to be addressed prior to commercial use. For example, hybrid perovskites were found to decompose when exposure to the ambient atmosphere due to the chemical reactions between themselves and humidity/oxygen in the air [6]. In this regard, various encapsulation technologies have been applied to hybrid perovskite photovoltaic devices to increase their service life [7]. Nevertheless, it should be noted that much less attention is paid to the chemical reactions between lead iodides and oxygen ions diffused from oxide substrates or electron transport layers, which migrate under the influence of light and temperature. This issue is addressed in this article, where the photochemical stability of glass/MAPbI 3 , ITO/MAPbI 3 , and TiO 2 /MAPbI 3 junctions was systematically studied with the assistance of X-ray photoelectron spectroscopy (XPS).

Experimental
Glass slides (ISOLAB objective glass), glass/ITO (5 Ω, Luminescence Technology Corp., New Taipei, Taiwan), and glass/TiO 2 (10 Ω, Luminescence Technology Corp., New Taipei, Taiwan) substrates were sequentially cleaned with toluene and acetone and sonicated in deionized water, acetone, and isopropanol. The 0.5 M MAPbI 3 precursor solutions in dimethylformamide (DMF) were spin-coated on top of the glass, glass/ITO, and glass/TiO 2 substrates at 4000 rpm inside a nitrogen-filled glove box. Toluene (100 µL) was dropped seven seconds after the start of the spinning thus quenching the precursor and inducing the film crystallization. Spinning was continued for 30 s and then the deposited films were annealed inside a glove box at 100 • C for 5 min. The photochemical aging experiments were performed under well-controlled anoxic conditions using specially designed setups integrated into the dedicated MBraun glove box using an LG sulfur plasma lamp as a standard light source, which provides a good approximation of the solar AM1.5G spectrum. The light power at the samples was~70 mW/cm 2 , while the temperature was 45 ± 2 • C (enabled by intense fan cooling of the sample stage).
Scanning electron microscopy images were obtained on Zeiss SUPRA 25 equipment (Jena, Germany). The samples were prefixed on a microscope stage inside the glove box to reduce the exposure time in the air down to~1 min. XRD patterns were recorded using a Bruker D8 instrument (Billerica, MA, USA).
XPS was used to measure core level and VB spectra with help of a PHI XPS 5000 VersaProbe spectrometer (ULVAC-Physical Electronics, Chanhassen, MN, USA) with a spherical quartz monochromator and an energy analyzer working in the range of binding energies from 0 to 1500 eV. The energy resolution was ∆E ≤ 0.5 eV. The samples were measured at a pressure of 10 −7 Pa. Finally, XPS spectra were processed using PHI MultiPak 9.9.0.8 software.

Results and Discussion
All samples showed notable aging after 300 h of light soaking even under anoxic conditions as could be concluded from the change of their visual appearance: the films become opaque and their color evolved from dark-brown to grayish-yellow. This conclusion was well supported by the scanning electron microscopy (SEM) images shown in Figure 1. Indeed, the typical grainy perovskite film morphology characteristic for the pristine samples changed dramatically after photochemical aging as could be concluded from the formation of poorly interconnected islands most probably represented by the aging products. Interestingly, the morphology of the aged films strongly depends on the type of substrate. Thus, the films grown on soda-lime glass showed the formation of big irregularly shaped islands decorated with tiny spherical or ellipsoidal particles. On the contrary, the perovskite films grown on ITO degraded with the formation of separate grains with uniform shape and 50-60 nm size. Furthermore, using TiO 2 as a substrate promoted the formation of some layered structures.
The evolution of the UV-vis absorption spectra shown in Figure 2 clearly illustrates that glass is the most inert substrate for MAPbI 3 since the photobleaching of the perovskite films occurs at the lowest rate. Furthermore, the absorption profile of the films after 200 h of aging closely resembles that of PbI 2 as could be concluded from the presence of the sharp feature at~500 nm. In contrast, the decomposition of MAPbI 3 films on ITO and TiO 2 goes much faster and the absorption profiles of the films become featureless after 200 h, which is suggesting the predominant formation of the metallic lead rather than PbI 2 .
Deeper insight into the chemistry of the light-induced decomposition of MAPbI 3 films deposited on different substrates was gained using X-ray diffraction (XRD). The perovskite films grown on glass showed the predominant formation of PbI 2 with just traces of metallic lead (Figure 3a), which is consistent with the UV-vis spectral data. On the contrary, the intensity of the Pb 0 peak became comparable or even higher than that of the major PbI 2 signature in the case of the films deposited on TiO 2 and ITO substrates (Figure 3b,c). However, one should keep in mind that the intensity of the peaks may not fully correlate with the actual phase composition of the samples due to e.g., different crystallinity of the components. Still, the obtained results support the conclusion that the substrate material affects not just the rate of the light-induced decomposition, but also the composition of the formed products.    Table 1. Measurements of the surface composition make it possible to determine their changes in the samples under study, while analysis of N:Pb and I:Pb ratios allow us to trace the photochemical degradation of MAPbI 3 depending on the exposure time to visible light. As can be seen in the sample of glass/MAPbI 3 , the N:Pb ratio exhibits a moderate decrease from 0.71 to 0.36 as the aging time gradually increases from 25 to 100 h, which indicates the slow, partial decomposition of the organic cation. Only at a 200-h exposure, the N 1s signal ceases to be recorded due to the complete evaporation of nitrogen species. Notably, the I:Pb ratio of MAPbI 3 does not exhibit a significant change (within 2.75-1.95) with the light-soaking time.   Surprisingly, much more significant changes in these ratios depending on aging time (25-200 h) were observed for the other two systems. For ITO/MAPbI 3 , the N:Nb ratio decreases from 0.73 to 0.13, and the I:Pb ratio decreased from 2.70 to 1.56 while, for TiO 2 /MAPbI 3 , the N:Nb ratio is reduced from 0.90 to 0.24 and the I:Pb ratio is reduced from 2.89 to 1.24. As shown earlier [8][9][10], a decrease in the N:Nb ratio indicates the decomposition of the organic cation, whereas reduction of I:Pb suggests the formation of PbI 2 , a product of decomposition of hybrid perovskites. On the one hand, these conclusions fit well into the degradation scheme as proposed in [11,12]: CH 3 NH 3 PbI 3 → PbI 2 solid + CH 3 NH 3 I → PbI 2 solid + CH 3 I gas + NH 3 gas (1) On the other hand, glass/MAPbI 3 turned out to be the most resistant to visible light irradiation as compared to ITO/MAPbI 3 and TiO 2 /MAPbI 3 .
Measurements of XPS N 1s, Pb 4d, and I 3d core-level spectra with a high energy resolution are capable to verify the validity of these conclusions. Relevant data are given in Figures 5-7. As first shown in Figure 5, the relative intensities of XPS N 1s spectra normalized to XPS Pb 4d-intensity show that some dependence on the aging time is still visible for the glass/MAPbI 3 system; however, it is not actually observed for the other two ones, for which the N 1s/Pb 4d relative intensity drops sharply starting from 50 h.   Figures 6 and 7 respectively present the dependence of the measured XPS Pb 4f and I 3d spectra of these three systems on the aging time and compare them with the spectra of PbI 2 compound that is the final product of photochemical degradation of MAPbI 3 hybrid perovskite. As shown, the energy position of XPS Pb 4f and I 3d spectra of PbI 2 is a benchmark, the proximity to which makes it possible to judge the degree of photochemical degradation. Notably, another benchmark is the energy position of the original (unirradiated) perovskites. The energy position of XPS Pb 4f and I 3d spectra measured under different aging times between these two reference points allows us to judge the degree of photochemical decomposition of the investigated materials. To better analyze XPS Pb 4f spectra, we have one more reference, which is the energy position of the pure lead metal (Pb 0 ). According to the degradation mechanism proposed in [13], further irradiation of PbI 2 (one of the primary degradation products) will lead to its decay into MAI and Pb 0 . Considering all of these factors, it is possible to analyze the experimental data presented in Figures 6 and 7 as follows. According to these data, XPS Pb 4f and I 3d spectra of irradiated glass/MAPbI 3 system are close in energy position to initial materials, which indicates their superior photochemical stability. On the contrary, for ITO/MAPbI 3 and TiO 2 /MAPbI 3 systems, a high-energy shift of XPS Pb 4f and I 3d spectra with the aging time is observed, and this effect is much more pronounced for the TiO 2 /MAPbI 3 system. In addition, for both systems, Pb 0 phase loss is also observed in XPS Pf 4f spectra and the relative intensity of this contribution is increasing with exposure time. This means that further decomposition of PbI 2 occurs in this case. Thus, the above results of high-energy resolved XPS N 1s, Pb 4f, and I 3d spectra leads to the same conclusion that we made from measurements of XPS survey spectra. That is, among these three studied systems, the glass/MAPbI 3 system is the most resistant to visible light irradiation, then ITO/MAPbI 3 and TiO 2 /MAPbI 3 , respectively, go in terms of the degree of stability loss.
We then try to understand the reason for the difference in photochemical stability of the same MAPbI 3 perovskite prepared on three different substrates. Oxygen is a common component for all three substrates used; therefore, first of all, the effect of irradiation on the state of oxygen in these systems is of interest. For this purpose, measurements of XPS O 1s spectra were carried out and the results are presented in Figure 8. As in the previous cases, the chemical state of the investigated element is herein determined by the energy position of the XPS spectrum. There are two starting points here: the chemical position of adsorbed and chemically bound oxygen (lattice oxygen). According to Ref. [14], the energy position of hydrogenated oxygen (O-H) is fixed in the range of 531.5-532.5 eV, and oxygen chemically bonded to the metal (O-Me) fluctuates in the range of 529-530.5 eV [15]. Hence, the data shown in Figure 5 allow us to conclude that, for the glass/MAPbI 3 system, the presence of only adsorbed oxygen on the sample surface is fixed. However, for ITO/MAPbI 3 system, the chemically bound oxygen already appears on the surface of the sample, and its concentration increases with the aging time. For TiO 2 /MAPbI 3 system, the contribution of chemically bound oxygen is already decisive on the surface of the sample. It should be noted that the most suitable metal for oxidation in these systems is lead and the next stage of our study is to obtain experimental data to support this judgment. For this purpose, we measured XPS valence band (VB) spectra of as-prepared (fresh) and light-soaked (at 200 h) glass/MAPbI 3 , ITO/MAPbI 3 , and TiO 2 /MAPbI 3 samples as shown in Figure 9. Identification of XPS VB Pb 6s and I 5p subbands in this figure is in accordance with the DFT calculation of the electronic structure of MAPbI 3 performed in Ref. [9]. First of all, we note the appearance of a new low-energy feature at the energy range of 6-9 eV in XPS VB spectra of ITO/MAPbI 3 and TiO 2 /MAPbI 3 , where, according to the DFT calculation of MAPbI 3 [9], Pb 6s states are concentrated, while this feature is absent in the spectrum of glass/MAPbI 3 system. The origin of these changes becomes clear from a comparison of these XPS spectra with the calculation result of the total density of states of lead oxide [16], which is also shown in Figure 9. The low-energy subband of the calculated spectrum located at 6-8 eV, which is the result of the hybridization of Pb 6s-O 2p states, correlates well in energy with this new feature of XPS VB spectra. Thus, from the correspondence between the experimental and calculated data, it can be concluded that lead atoms in ITO/MAPbI 3 and TiO 2 /MAPbI 3 systems chemically interact with oxygen atoms. It should be noted that, in XPS VB spectra, an increased contribution of Pb 6s-O 2p states is observed for TiO 2 /MAPbI 3 system compared to the ITO/MAPbI 3 system, which is in good agreement with the change in the contributions of chemically bound oxygen (O-Me) to the XPS O 1s spectra ( Figure 8).

Conclusions
Measurements of XPS spectra have shown that, under irradiation of MAPbI 3 hybrid perovskites with visible light (25-200 h), their partial photochemical degradation occurs and it is more pronounced on ITO and TiO 2 substrates than on a glass substrate. This degradation is accompanied by the appearance of MAI, PbI 2 , and Pb 0 decay products as evidenced by XPS N 1s, Pb 4f, and I 3d spectra. At the same time, the oxygen ions migrate from the substrates to perovskite and interact with the decomposition products to result in the formation of OH-containing and lead-oxygen species as concluded from the XPS O 1s and VB spectra. The oxidation of the perovskite material by the oxygen from the substrate induces additional photochemical degradation, and, therefore, both ITO/MAPbI 3 and TiO 2 /MAPbI 3 junctions are found to be less stable to light soaking than glass/MAPbI 3 system.