In-Situ Nano-Auger Probe of Chloride-Ions during CH3NH3PbI3−xClx Perovskite Formation

Organo-halide perovskite solar cells (PSCs) have emerged as next-generation photovoltaics, owing to their high power-conversion efficiency (PCE), lower production cost, and high flexibility. ABX3-structured methylammonium lead triiodide (CH3NH3PbI3 or MAPbI3) perovskite is a widely studied light-absorbing material in PSCs. Interestingly, a small amount of chlorine incorporation into MAPbI3 increases charge carrier diffusion lengths (from 129 nm to 1069 nm), which enables planar structured PSCs with high PCEs. However, existence of chloride ions in the final perovskite film is still under debate. Contrastingly, few studies reported a negligible amount or absence of chloride ions in the final film, while others reported detection of chloride ions in the final film. Herein, we observed the microstructure and chlorine content of MAPbI3−xClx thin films with increasing temperature via an in-situ nano-Auger spectroscopy and in-situ scanning electron microscopic analysis. The relative precipitation of MAPbI3−xClx films occur at lower temperature and MAPbI3−xClx grains grow faster than those of MAPbI3 grains. Local concentrations of chlorine at intragrain and the vicinity of grain boundary were analyzed to understand the behavior and role of the chloride ions during the microstructural evolution of the MAPbI3−xClx films.


Introduction
Organo-halide perovskite (OHP) materials are a hot topic in photovoltaic research due to their excellent light harvesting ability and relatively lower processing costs [1][2][3]. In the past few years, perovskite solar cells (PSCs) have seen tremendous uplift in certified power conversion efficiencies (PCE), less than 14% in 2013 to >25% in 2020 [4][5][6]. The archetypal ABX 3 -structured methylammonium lead halides such as MAPbI 3 and MAPbI 3−x Cl x are the most intensely studied reference materials in PSCs, owing to their remarkable optoelectronic properties [7,8]. MAPbI 3 has large light-absorption coefficients (~10 5 cm −1 at 550 nm) due to a direct bandgap (~1.55 eV) between the valence (Pb s and I p orbital) and the conduction (Pb p orbital) band edge where intramolecular charge transfer occurs between the high-density orbitals [9,10]. Importantly, carrier diffusion length and band gap of MAPbI 3 are controlled by compositional engineering. The fine control over optoelectronic properties and ease of synthesis has driven MAPbI 3 to broad applications, including PSCs [11], solar-water splitting [12], lasing [13], and light-emitting diodes [14]. In addition, the incorporation of a small amount of bromine or chlorine into MAPbI 3 is known to improve device properties [15]. Since the first report on PSCs using mixed halide MAPb 3−x Cl x as a light absorber, the PCE of the MAPb 3−x Cl x devices has already exceeded 18% [16,17]. The mixed halide MAPbI 3−x Cl x was found to have long carrier diffusion length of 1µm, about one order of magnitude longer than pure MAPbI 3 system and 2.5 times higher diffusion constants (from 0.017 cm 2 s −1 to 0.042 cm 2 s −1 ) compared to pure MAPbI 3 [18]. Despite many investigations, the role of chloride ions in MAPbI 3−x Cl 3 mixed-halide perovskites has not been rationalized [19][20][21][22]. Understandably, the ambiguity about chloride ions in MAPbI 3−x Cl x films is due to inefficient analytical techniques. As such, the concentration of chloride ions in MAPbI 3−x Cl x films after the annealing process is below the detection limit of an energy dispersive spectrometer (EDS) and X-ray photoelectron spectroscopy (XPS). Hence, to examine the role and volatilization behavior of chloride ions, it is desirable to have in-situ measurement techniques that are better than the conventional scanning electron microscopy (SEM) or EDS.
Herein, we use in-situ SEM and high-resolution in-situ nano-Auger electron spectroscopy (n-AES) to explore the role of chloride ions during MAPbI 3−x Cl x thin film formation. The surface microstructures were analyzed by in-situ SEM, quantitative components were analyzed using in-situ n-AES, and the structural analysis was performed using the X-ray diffraction (XRD) technique. The microstructural analysis was carried out by increasing the annealing temperature in-situ to confirm the volatilization behavior of chloride ions with increasing temperature. To quantitate chloride ions in MAPbI 3−x Cl x thin films, samples were in-situ annealed at a constant temperature over a period of time. During the volatilization, a nano-sized, high-resolution in-situ nano-auger system was used to quantitatively analyze the local concentration of chloride ions between the grain interior and grain boundary. The microstructure, grain growth, and diffusion or sublimation behavior of chloride ions during MAPbI 3−x Cl x thin film formation are succinctly discussed.

Preparation of TiO 2 Blocking Layer
All the samples were prepared on TiO 2 blocking layer-coated Silicon (n-type, resistivity = 0.007~0.002 Ω·cm, LG Siltron) substrates. The TiO 2 blocking layer was prepared via the spin-coating of a 0.15 M titanium (IV) diisopropoxide bis(acetylacetonate) (75 wt.%) in 1-butanol (Sigma-Aldrich, 99.8%) on Si substrates. The TiO 2 blocking layer precursor solution was spin-coated at 3000 rpm for 20 s followed by soft baking at 125 • C for 5 min and hard baking at 500 • C for 30 min.

Formation of MAPbI 3−x Cl x Layer
The MAPbI 3−x Cl x precursor solution was prepared with 1:1:1 molar ratio of MAI, PbI 2 +PbCl 2 , and DMSO in DMF solvent at room temperature. The chlorine was added at 0.01 at% and 0.10 at%. Here, at% means atomic percentage of "X" cite in the perovskite. The MAPbI 3−x Cl x precursor solution was spin-coated at 4000 rpm for 20 s. To correlate our nano-AES analysis with the latest developments, we used an antisolvent method to fabricate a high-quality perovskite layer. During the spin coating, 500 µL of DEE was dropped on to the MAPbI 3−x Cl x solution. After the spin-coating, all the samples were soft-baked at 65 • C for 1 min. For the in-situ study, soft-baked samples were installed in an in-situ system. For other analyses, soft-baked samples were hard-baked at 130 • C for 10 min under vacuum (<10 −3 Torr).

Characterization
The structural properties were measured by XRD (X'pert, Pan analytical, Amelo, The Netherlands). The microstructural and the thickness of the thin films were measured using filed-emission scanning electron microscopy (FE-SEM; JEOL, JSM-6701F, Tokyo, Japan). In-situ microstructural imaging was conducted by conventional SEM (Zeiss, Gemini Supra 55 VP-SEM, Oberkochen, Germany) equipped with a home-built heating stage. Nano-AES experiments were carried out using an Oxford/Omicron Nano-Auger system furnished with an ultra-high vacuum chamber (base pressure~10-10 Torr), a field emission source, and a multi-channel hemisphere energy detector, as well as a heating specimen holder. The electron beam which can be focused to~10 nm diameter enabled probing chemical composition in the nanoscale. In-situ analyses were conducted during the ramping up of the temperature from room temperature to 130 • C (~20-50 • C/min) and maintaining at the temperature for 1 h, under vacuum (<10 −6 Torr).

Results and Discussion
Perovskite layers with a composition of MAPbI 3−x Cl x were prepared on TiO 2 layers to make similar conditions for the grain growth of perovskite film that are used for optoelectronic devices, as the surface state of the substrates significantly affects the grain growth behavior. To observe the diffusion or sublimation of chloride ions, SEM analysis was performed in-situ while increasing the baking temperature. Figure 1A,B shows the in-situ SEM images of MAPbI 3−x Cl x (where x = 0, 0.10) samples recorded at various growth temperatures. The soft baking was performed at 65 • C to remove the solvents or anti-solvents and nucleate the perovskite crystallites. As shown in Figure 1B, at 65 • C MAPbI 3−x Cl x samples (where x = 0.1) displayed a larger grain size compared to the chlorine-free sample. This is not the difference in grain size during the hard-backing process for grain growth, but the difference in grain size during the soft-baking process. Additionally, soft baking was also performed at a 45 • C, and the result was the same as soft baking at 65 • C, suggesting chloride-induced grain growth. During the hard baking of MAPbI 3−x Cl x samples at a higher temperature, precipitates began to appear (85 • C under Figure 1A,B). In the case of x = 0 composition, precipitation started at about 90 • C, but in the case of x = 0.1 the composition precipitation started at about 80 • C. This means that the chlorine-related compound was volatilized at lower temperature. When increasing baking temperature above precipitation temperature (130 • C under Figure 1A,B), regardless of composition (x = 0, 0.1) the grains started to converge as a continuous network with the generation of pin holes. As shown in Figure 1C, the grain size difference was larger at the lower temperature, and the difference in grain size was lower at the higher temperature, asserting chlorine-induced grain growth even at lower temperatures. This phenomenon can be explained by the difference in phase formation energy (E f ). When the E f is low, a phase is formed at a lower temperature and the remaining energy can be used for grain growth. The phase formation energy can also be explained by the difference of the phase formation enthalpy (H f ). The H f of MAPbI 3 is 34.50 ± 1.10 kJ/mol, and the H f of MAPbCl 3 is −9.03 ± 1.68 kJ/mol [23,24]. According to the first law of thermodynamics, ∆G = ∆H − T∆S, where T∆S is negligible in the experimental conditions, so ∆G ≈ ∆H. A negative Gibbs-free energy value means a spontaneous reaction and a stable state. Thus, the MAPbCl 3 phase is more stable thermodynamically. Therefore, it can be deduced that the more chlorine is added to MAPbI 3 , the more thermodynamically stable it is and the E f becomes lower. more chlorine is added to MAPbI3, the more thermodynamically stable it is and the Ef becomes lower. The crystallization behavior and phase formation of MAPbI3−xClx thin films was evaluated by recording XRD patterns. Figure 1D shows the XRD data of MAPbI3−xClx (x = 0, 0.1) samples soft baked at 65 °C. After soft baking, it was confirmed that the MAPbI3 phase was formed in all the compositions regardless of the chlorine amount. A closer observation reveals that the (002) and (110) planes in MAPbI3−xClx with x = 0.1 composition were larger than pure MAPbI3. This is in good agreement with the grain growth observed in SEM analyses (65 °C under Figure 1A,B). Besides, intermediate phase diffraction peaks of PbI2-DMSO and PbCl2-DMSO were seen at lower 2θ angles < 10° [25]. The MAPbI3 phase was formed because chlorine was added in an amount smaller than the solubility limit. The calculated and experimental solid solubility limit of chlorine in the MA site of MAPbI3 is about 4 at% [26,27]. Therefore, chlorine can be present, as the composition x = 0.10 is about 3.3 at%. Figure 1E shows the XRD data of a sample obtained by hard baking MAPbI3−xClx (x = 0, 0.1) thin films at 130 °C. After hard baking at 130°C, the diffractions from the intermediate (PbI2/PbCl2-DMSO) phase disappeared in both of the MAPbI3−xClx (x = 0, 0.1) compositions and a no-chlorine-related phase was observed in both the samples, consistent with the reported literature [28]. Interestingly, intensities of all diffraction peaks of x = 0.1 composition were higher than those of the chlorine-free composition (x = 0), suggesting larger grain growth, as illustrated in the SEM analysis ( Figure 1A,B).
To observe the volatilization behavior of chloride ions, SEM analysis was performed in-situ while increasing the baking time at a fixed temperature of 130 °C. Figure 2A  The crystallization behavior and phase formation of MAPbI 3−x Cl x thin films was evaluated by recording XRD patterns. Figure 1D shows the XRD data of MAPbI 3−x Cl x (x = 0, 0.1) samples soft baked at 65 • C. After soft baking, it was confirmed that the MAPbI 3 phase was formed in all the compositions regardless of the chlorine amount. A closer observation reveals that the (002) and (110) planes in MAPbI 3−x Cl x with x = 0.1 composition were larger than pure MAPbI 3 . This is in good agreement with the grain growth observed in SEM analyses (65 • C under Figure 1A,B). Besides, intermediate phase diffraction peaks of PbI 2 -DMSO and PbCl 2 -DMSO were seen at lower 2θ angles < 10 • [25]. The MAPbI 3 phase was formed because chlorine was added in an amount smaller than the solubility limit. The calculated and experimental solid solubility limit of chlorine in the MA site of MAPbI 3 is about 4 at% [26,27]. Therefore, chlorine can be present, as the composition x = 0.10 is about 3.3 at%. Figure 1E shows the XRD data of a sample obtained by hard baking MAPbI 3−x Cl x (x = 0, 0.1) thin films at 130 • C. After hard baking at 130 • C, the diffractions from the intermediate (PbI 2 /PbCl 2 -DMSO) phase disappeared in both of the MAPbI 3−x Cl x (x = 0, 0.1) compositions and a no-chlorine-related phase was observed in both the samples, consistent with the reported literature [28]. Interestingly, intensities of all diffraction peaks of x = 0.1 composition were higher than those of the chlorine-free composition (x = 0), suggesting larger grain growth, as illustrated in the SEM analysis ( Figure 1A,B).
To observe the volatilization behavior of chloride ions, SEM analysis was performed in-situ while increasing the baking time at a fixed temperature of 130 • C. Figure 2A,B shows the in-situ SEM images of MAPbI 3−x Cl x (x = 0, 0.1) thin films annealed at 130 • C for various (0, 5, and 30 min) time intervals. The top-view surface microstructure was observed to be same irrespective of the chlorine amount. However, it was seen that long-term annealing resulted in pin-hole formation in the x = 0.10 composition at a faster rate than x = 0, owing to chloride ion sublimation. Figure 2C shows the grain size of MAPbI 3−x Cl x (x = 0, 0.1) thin films by in-situ annealing. After 5 min of annealing at 130 • C, the grain size of MAPbI 3−x Cl x (x = 0.1) was comparatively larger than MAPbI 3−x Cl x (x = 0). This is because at a higher temperature the diffusion of chloride ions is much faster than iodide ions, resulting in more space for larger grain growth. Indeed, long-term baking converges large grains into a long network with the formation of precipitates near the grain boundaries. The evaporation of chloride ions and the presence of precipitates can be understood by XRD data in Figure 2D,E. Besides, after long-time annealing (30 min) at 130 • C, XRD results of MAPbI 3−x Cl x (x = 0, 0.1) thin films show PbI 2 formation. In the case of x = 0.1 composition, PbI 2 formation was observed at a shorter time than the chlorine-free composition (5 min data under Figure 2E), and after 30 min of baking the signature peak (002) intensity of x = 0.1 was lower than x = 0. The precipitation of PbI 2 at a lower annealing time in x = 0.1 means that methylammonium (MA) was deficient and the chlorine was volatilized with MA, i.e., the MACl gas phase [29]. Hence, PbI 2 was precipitated through grain boundaries, implying that the chlorine-related phase could be expected to evaporate through grain boundaries. To confirm this, n-AES analysis was performed in-situ to determine the difference in chlorine concentrations at the grain interior and grain boundary. various (0, 5, and 30 min) time intervals. The top-view surface microstructure was observed to be same irrespective of the chlorine amount. However, it was seen that longterm annealing resulted in pin-hole formation in the x = 0.10 composition at a faster rate than x = 0, owing to chloride ion sublimation. Figure 2C shows the grain size of MAPbI3−xClx (x = 0, 0.1) thin films by in-situ annealing. After 5 min of annealing at 130 °C, the grain size of MAPbI3−xClx (x = 0.1) was comparatively larger than MAPbI3−xClx (x = 0). This is because at a higher temperature the diffusion of chloride ions is much faster than iodide ions, resulting in more space for larger grain growth. Indeed, long-term baking converges large grains into a long network with the formation of precipitates near the grain boundaries. The evaporation of chloride ions and the presence of precipitates can be understood by XRD data in Figure 2D,E. Besides, after long-time annealing (30 min) at 130 °C, XRD results of MAPbI3−xClx (x = 0, 0.1) thin films show PbI2 formation. In the case of x =0.1 composition, PbI2 formation was observed at a shorter time than the chlorine-free composition (5 min data under Figure 2E), and after 30 min of baking the signature peak (002) intensity of x = 0.1 was lower than x = 0. The precipitation of PbI2 at a lower annealing time in x = 0.1 means that methylammonium (MA) was deficient and the chlorine was volatilized with MA, i.e., the MACl gas phase [29]. Hence, PbI2 was precipitated through grain boundaries, implying that the chlorine-related phase could be expected to evaporate through grain boundaries. To confirm this, n-AES analysis was performed in-situ to determine the difference in chlorine concentrations at the grain interior and grain boundary.   Figure 3A displays the quantitative change in the chlorine amount of representative MAPbI 3−x Cl x (x = 0.1) samples at the outermost surface with baking time given by conventional AES measurements. When baking was performed at 130 • C, we found that a large amount of chlorine had volatilized following 30 min of baking time. Then, a few atomic layers were removed from the outermost surface by gentle Ar ion milling to clearly observe the compositional difference of the grain interior and grain boundary. The reduced chlorine is attributed to evaporation of MACl, as the evaporation temperature of is about 100 • C [29,30]. Figure 3B shows a typical in-situ high-resolution SEM image under the n-AES mode where the e-beam current is much higher than that for standard SEM imaging, with the marks pointing to the regions where Auger signals were collected under the focused beam. Figure 3C illustrates the quantitative difference in the amount of chlorine between grain interior and grain boundary of MAPbI 3−x Cl x (x = 0.1) samples before/after annealing at 130 • C for 10 min. Before annealing, the as-formed MAPbI 3−x Cl x (x = 0.1) film exhibited higher concentration of chloride ions in the grain interior than grain boundaries. After gentle annealing at 130 • C for 10 min, grain interior chloride ions diffused out to grain boundaries via chlorine-to-iodine switching. This is likely because the ionic radius of chlorine is lower than the iodine, which permits intra-grain ions interdiffusion. Figure 3D shows the average chlorine percentage of MAPbI 3−x Cl x (x = 0.1) film in the solution, outermost surface, grain interior, and at the grain boundaries as a function of annealing time. The overall chlorine percentage in the as-grown MAPbI 3−x Cl x (x = 0.1) film was 1.5%, which is similar to the precursor solution (1.67%). The overall chlorine content was gradually decreased to 0.15% when annealed up to 30 min. This clearly implies that most of chloride ions were removed from MAPbI 3−x Cl x (x = 0.1) by evaporation in the form of MACl gas. Additionally, the chloride ions at the grain interior and grain boundaries also gradually decreased with longer annealing time. Although overall chlorine content was volatilized with an increase in annealing time, the chlorine content at the grain interior and grain boundaries was reversed after 5 min, as shown in Figure 3E. The ratio of chlorine composition difference at the grain interior and grain boundaries was less than 30% and it remained constant until 10 min of annealing time. This implies that the chloride-ion diffusion at a lower annealing time plays an important role in the grain growth of MAPbI 3−x Cl x (x = 0.1) thin films, which is in good agreement with XRD and SEM grain growth analyses ( Figure 2). Importantly, a small fraction of chloride ions still remained at grain boundaries even after the annealing process of 30 min. This small fraction of inherent chlorine content is expected to be one of reasons for large grain growth and improved diffusion coefficient of mixed halide MAPbI 3−x Cl x compared to pure MAPbI 3 composition. In support of previous reports [31,32], the in-situ experiment results presented here further suggest that the presence of chlorine precursors in MAPbI 3 thin film formation may play an important role in grain growth, which is closely relevant to engineering morphology and electrical properties that are desirable for photovoltaic devices.

Conclusions
In summary, the incorporation of a small amount of chlorine increases grain size of resultant MAPbI3−xClx thin films compared to pure MAPbI3. The mixed halide MAPbI3−xClx phase is thermodynamically more stable and occurs at a lower formation energy than MAPbI3. Microstructure analyses by in-situ n-AES suggest that chloride ions inside the grain interior diffuse out to grain boundaries prompting faster grain growth of MAPbI3−xClx thin films. At a lower baking temperature, the chlorine gradient is high at the grain interior, and when the MAPbI3−xClx phase is baked for a long time, it changes to an MAPbI3-like microstructure due to the volatilization of chloride ions. Although the XRD results could not prove the chloride content in the MAPbI3−xClx phase, the in-situ n-AES study concluded that a small fraction of chlorine was still present even after the longer baking time.

Conclusions
In summary, the incorporation of a small amount of chlorine increases grain size of resultant MAPbI 3−x Cl x thin films compared to pure MAPbI 3 . The mixed halide MAPbI 3−x Cl x phase is thermodynamically more stable and occurs at a lower formation energy than MAPbI 3 . Microstructure analyses by in-situ n-AES suggest that chloride ions inside the grain interior diffuse out to grain boundaries prompting faster grain growth of MAPbI 3−x Cl x thin films. At a lower baking temperature, the chlorine gradient is high at the grain interior, and when the MAPbI 3−x Cl x phase is baked for a long time, it changes to an MAPbI 3 -like microstructure due to the volatilization of chloride ions. Although the XRD results could not prove the chloride content in the MAPbI 3−x Cl x phase, the in-situ n-AES study concluded that a small fraction of chlorine was still present even after the longer baking time.