Structural characterization of bulk and nanoparticle lead halide perovskite thin films by (S)TEM techniques

Lead halide (APbX3) perovskites, in polycrystalline thin films but also perovskite nanoparticles (NPs) has demonstrated excellent performance to implement a new generation of photovoltaic and photonic devices. The structural characterization of APbX3 thin films using (scanning) transmission electron microscopy ((S)TEM) techniques can provide valuable information that can be used to understand and model their optoelectronic performance and device properties. However, since APbX3 perovskites are soft materials, their characterization using (S)TEM is challenging. Here, we study and compare the structural properties of two different metal halide APbX3 perovskite thin films: bulk CH3NH3PbI3 prepared by spin-coating of the precursors in solution and CsPbBr3 colloidal NPs synthetized and deposited by doctor blading. Both specimen preparation methods and working conditions for analysis by (S)TEM are properly optimized. We show that CH3NH3PbI3 thin films grown by a one-step method are composed of independent grains with random orientations. The growth method results in the formation of tetragonal perovskite thin films with good adherence to an underlying TiO2 layer, which is characterized by a photoluminescence (PL) emission band centered at 775 nm. The perovskite thin films based on CsPbBr3 colloidal NPs, which are used as the building blocks of the film, are preserved by the deposition process, even if small gaps are observed between adjacent NPs. The crystal structure of CsPbBr3 NPs is cubic, which is beneficial for optical properties due to its optimal band gap. The absorption and PL spectra measured in both the thin film and the colloidal solution of CsPbBr3 NPs are very similar, indicating a good homogeneity of the thin films and the absence of aggregation of NPs. However, a particular care was required to avoid long electron irradiation times during our structural studies, even at a low voltage of 80 kV, as the material was observed to decompose through Pb segregation.


Introduction
Lead halide (APbX3) perovskites are highly promising materials for optoelectronic and photonic devices. In fact, these materials have been widely used in solar cell [1][2][3], light-emitting diode [4][5] and lasing [6][7], because of their high photo-conversion efficiencies (up to 22.7%), excellent light emission properties [8], and low cost fabrication [1]. Their outstanding properties include high absorption coefficients and sharp absorption edges [9], band gaps that are tunable with composition, and high emission quantum yields at room temperature. Hybrid CH 3 NH 3 PbI 3 has attracted particular attention, as it is the first lead halide perovskite used in photovoltaic applications [9][10]. CH 3 NH 3 PbI 3 semiconductor has a direct band gap of ∼1.6 eV and high carrier mobility, resulting in solar cells with conversion efficiencies of up to 22.7% [11], and optical amplifiers with low thresholds of stimulated emission [12]. However, CH 3 NH 3 PbI 3 films degrade under ambient conditions, especially in the presence of moisture [13][14]. In contrast, inorganic perovskites have demonstrated higher stability [15][16]. In particular, CsPbBr 3 nanocrystals have been shown to exhibit ∼90% photoluminescence (PL) efficiency with a narrow full width at half maximum (FWHM) [17][18] and have a highly emissive crystal phase that is stable at room temperature [19], leading to enormous potential for applications in optoelectronic devices. Examples of inorganic perovskite-based devices are light-emitting diodes [20], photodetectors [21] and solar cells [22].
Although both bulk and nanoparticle (NP) thin films have similar perovskite lattices, they have different structural properties that need to be understood to optimize their optoelectronic properties and to integrate them efficiently in solid state devices. Control of their structural properties, such as crystallinity [23] and composition [24], is a key issue for device efficiency. Crystalline defects and grain boundaries can act as non-radiative recombination sites that affect photovoltaic properties [25][26]. Parameters such as film roughness and surface coverage after deposition of the material on a substrate also need to be controlled to obtain smooth and continuous layers in high-efficiency solar cells [27][28]. Other issues include the long-term stability of the active layer. The coverage of the perovskite with polymeric materials is one of the most common solutions to improve the stability of the perovskites. Thus, polymers such as poly(methyl methacrylate) (PMMA), polyvinyl alcohol or poly vinylpyrrolidone have been considered. Among them, PMMA has shown to provide larger air stability to the perovskite [29]. FTIR has shown that this could be due to the interaction between the carbonyl groups of PMMA and the Pb cations of the lead halide perovskite [30]. Additionally, the PMMA coverage also benefits the propagation of the excitation beam along the semiconductor [12].
Detailed information about the structural properties of the material after deposition on a substrate is therefore essential. However, only indirect measurements are typically used. For example, the defect density in CH 3 NH 3 PbI 3 perovskites has been quantified using first-principles calculations and spectroscopic measurements [31]. Direct measurements of the structural properties of APbX 3 perovskites are required to understand and improve synthesis, deposition processes and the influence of the A cation component in the electronic band structure and the subsequent optical and optoelectronic properties. For this purpose, transmission electron microscopy (TEM) and (scanning) transmission electron microscopy ((S)TEM) are powerful techniques for the local microstructural characterization of advanced materials [32][33]. Crystalline structure [34] and iodide migration [35] have been studied in CH 3 NH 3 PbI 3 perovskites using these techniques. Furthermore, they have been used to identify new phases in CsPbBr 3 NPs, such as CsPb 2 Br 5 during the synthesis of CsPbBr 3 [36]. Although these studies demonstrate their benefit for the knowledge advance of perovskite materials, the characterization of lead halide perovskites by (S)TEM techniques presents particular challenges, as the incident electron beam can degrade the material. As this previous examples highlight, (S)TEM characterization of perovskites is not straightforward neither in standard layers nor in films formed by NPs, here we report, as far as we know, the first comparative study of these two important families of perovskite materials.
We analyze the structural properties of two different perovskite thin films, comprising bulk CH 3 NH 3 PbI 3 and CsPbBr 3 NPs, using (S)TEM techniques. Both films were deposited on SiO 2 /TiO 2 substrates, in order to ensure an easy cleavage, and covered by PMMA to enhance their stability to ambient conditions. Specimen preparation and imaging conditions for electron microscopy were optimized to avoid degradation of the materials during the measurements. Special attention was paid to understand the morphologies of the layers and their crystal structures. For this purpose, (S)TEM characterization of both kinds of perovskite thin films was properly compared in order to understand the influence of the composition and nature of the base material (chemical precursors or colloidal NPs) in the crystal structure and morphology of the resulting film. Since these parameters affect the functionalities of potential optoelectronic devices, we believe that this characterization could be beneficial in the implementation of future perovskite-based devices.

Bulk CH 3 NH 3 PbI 3 perovskite thin film preparation
Si/SiO 2 wafers were cleaned sequentially using acetone, ethanol and isopropanol. The wafers were dried using a N 2 stream and heated for 1 min at 100°C over a hot plate. A ∼40 nm thick layer of TiO 2 was deposited on the Si/SiO 2 wafer by spin-coating an anhydrous ethanol solution of titanium isopropoxide and heating at 500°C for 30 min in room atmospheric conditions. The lead halide perovskite layer was deposited in a glove-box by spin-coating 100 μl of the perovskite precursor DMF solution of 40% or 10% (w/w) onto the wafer. After deposition, the substrate was kept at room temperature for 15 min and then heated to 100°C for 1 h in an oven in air. After depositing the perovskite onto the substrate, a solution of PMMA in toluene was spin coated onto the sample and heated at 150°C.

CsPbBr 3 NP synthesis and sample deposition
In order to synthesize CsPbBr 3 NPs, PbBr 2 (0.4 mmol) and CsBr (0.4 mmol) were dissolved in DMF (10 ml). Upon complete dissolution of the precursors, oleic acid (1.0 ml) and oleylamine (0.5 ml) were added to stabilize the precursor solution. Then, the precursor solution (1.0 ml) was added dropwise to toluene (10 ml) under vigorous stirring. Strong green emission was observed immediately after injection [18]. Thin films of CsPbBr 3 NPs were prepared from the colloidal solution using a doctor blading technique. This deposition method ensures good film homogeneity and avoids agglomeration of the NPs, as we demonstrated previously for colloidal NPs [37]. Characterization PL measurements were performed in backscattering geometry using a continuous wave excitation laser at 404 nm under a weak pump fluency of 20 W cm −2 and collecting the backemitted PL into an HR4000 Ocean Optics spectrograph.
Absorption was obtained using a Shimadzu UV-2501PC spectrophotometer (UV-visible range), or by measuring the transmittance using a commercial reflectometer (Nanocalc from Micropack).
Electron-transparent specimens of CH 3 NH 3 PbI 3 were prepared for (S)TEM analysis using two methods. The first method involved conventional grinding, polishing and Ar + ion milling using a precision ion polishing system (PIPS) from Gatan. The second method involved focused ion beam (FIB) milling using a Helios Nanolab 650 DualBeam from FEI Company. The CsPbBr 3 specimen was also prepared using FIB milling. (S)TEM measurements were performed using an FEI Titan Themis Cubed TEM and an FEI Titan G2 80-200 ChemiSTEM equipped with a Super-X energy dispersive x-ray (EDX) spectrometer. Eje Z software [38] was used for the simulation of digital diffraction patterns (DDPs) of the perovskites. The CH 3 NH 3 PbI 3 perovskites considered are orthorhombic (Pnma), tetragonal (I4/mcm) and cubic (Pm3m), while the CsPbBr 3 perovskites considered are cubic (Pm3m), tetragonal (P4/mbm) and orthorhombic (Pnma).

Results and discussion
Figure 1(a) shows absorbance and PL spectra recorded from the CH 3 NH 3 PbI 3 perovskite film as blue and red lines, respectively. The sharp absorption edge is localized at 750-760 nm, while the exciton absorption peak is centered at 750 nm, in accordance with previous publications [9]. The PL emission band has a Gaussian shape centered on 775 nm which is associated with a tetragonal structure [39], and has a FWHM of 46 nm. The absorption of colloidal CsPbBr 3 NPs in this work (solid blue line in figure 1(b)) is associated with a band edge at ∼510 nm, with the excitonic peak localized at 501 nm, in agreement with previous results reported for this material [40]. The PL spectrum has a Gaussian shape (green solid line in figure 1(b)), which is centered at 517 nm, is related to the cubic structure [41] and has a FWHM of 21 nm. This red Stokes shift can be explained by aggregation of the NPs [42]. The film of colloidal NPs shows similar absorption and PL as the colloidal solution (blue and green dashed lines in figure 1(b), respectively), indicating good homogeneity and the absence of aggregation of NPs in the film.
TEM analysis of CH 3 NH 3 PbI 3 initially focused on the specimen prepared using a conventional method. Figure 2(a) shows a TEM image, in which the different layers of the structure can be distinguished. In order to study the crystalline quality of the layers, fast Fourier transform (FFT) analysis was carried out. Surprisingly, most of the material formed on the TiO 2 layer was found to be non-crystalline, as shown by characteristic amorphous rings in the corresponding FFT (see inset to figure 2(a)). EDX analysis shows that the composition of the amorphous material is consistent with that expected for the perovskite layer, as shown in the second inset to figure 2(a). The large amount of amorphous material present suggests that amorphization may have taken place during specimen preparation for TEM. FIB milling was therefore used as an alternative specimen preparation procedure. Figure 2(b) shows a TEM image of CH 3 NH 3 PbI 3 prepared using FIB milling, in which the perovskite and TiO 2 layers can now be observed. The FFT (shown as an inset) contains spots corresponding to a crystalline structure. A second inset shows a high-resolution TEM (HRTEM) image of CH 3 NH 3 PbI 3 , in which atomic planes can be distinguished, confirming the crystalline nature of the material. The advantage of FIB milling for preparation of halide perovskites for (S)TEM results from the use of a Ga + ion gun with a wellcontrolled beam to thin the specimen to electron transparency. The angle of incidence of the Ga + ions is almost parallel to the surface of the material during thinning, whereas in the PIPS used for conventional preparation the Ar + ion beam has an angle of 3.5°. Although the same ion voltage (∼2 kV) is used in the last step of the thinning process in both methods, the PIPS is more aggressive than the FIB for the present materials, despite higher voltage used for FIB milling during the intermediate steps of the process.
After optimization of the specimen preparation method, the conditions for TEM analysis need to be considered, including the choice of microscope accelerating voltage. Figures 2(c) and (d) show two TEM images of the same region of CH 3 NH 3 PbI 3 before and after analysis at 200 kV. The area analyzed (marked with a red rectangle) suffered clear beam damage during TEM examination. Radiolysis and knock-on damage are the two most significant interactions between an incident electron beam and a specimen in the TEM. In general, radiolysis is reduced at higher accelerating voltages, whereas knock-on damage is increased [43]. Damage to the specimen depends on the nature of the material. Radiolysis remains a major limitation when studying certain ceramics and minerals and most polymers [44]. In contrast, knock-on damage usually has a stronger effect on materials such as certain semiconductors and low-Z metals/ alloys [44]. Similar materials may behave differently. For example, the threshold energy required to produce a Frenkel defect by knock-on damage is smaller for Ge than for Si [45].
Our results suggest that the use of a high accelerating voltage is not appropriate for TEM studies of CH 3 NH 3 PbI 3 perovskites, as displacement damage seems to be significant. We therefore analyzed this specimen at a lower accelerating voltage of 80 kV. Figures 2(e) and (f) show TEM images of the same region of the CH 3 NH 3 PbI 3 perovskite before and after examination at 80 kV. The large and rapid damage to the material that occurred during the electron irradiation at 200 kV is no longer observed at 80 kV, demonstrating that the accelerating voltage has to be chosen carefully to avoid degradation of CH 3 NH 3 PbI 3 perovskite film. In the literature, the use of accelerating voltages of 100 kV [44][45][46] and 60 kV [47] and short exposure times [48] has been reported for the study of CH 3 NH 3 PbI 3 perovskites. Figure 3(a) shows a scanning electron microscopy (SEM) image of a CH 3 NH 3 PbI 3 lamella, in which different layers of the sample structure can be observed, recorded in the FIB workstation after specimen preparation. EDX spectroscopy was used to identify the layers (see inset). The perovskite layer does not have a uniform thickness because the layers are formed from the crystallization of grains that grow independently with different orientations and sizes. It has been reported that the one-step synthesis-deposition method that was used in the fabrication of the material is affected by an imbalance between nucleation and growth rates, leading to different morphologies of the perovskite layer [49]. A DMF intermediate was used during fabrication, as it has been reported to produce an enhancement [49] that contributes to control of the crystallization of the perovskite layer [50]. However, even after using the DMF intermediate, our results indicate that the layer thickness is not uniform. In order to investigate the initial steps of layer formation, a second sample, in which the concentration of the precursor in solution was four times lower, was analyzed.  shows an SEM image of the resulting structure, in which the perovskite has not crystallized in a uniform layer, but as islands of thickness 80-100 nm. The layer shows irregular growth from the initial steps of material deposition. As shown in the regions marked by squares, the initial perovskite layer is not always flat, but contains some uneven regions, which will affect the geometry of the material when a full layer is grown. The analysis allows the adherence of the perovskite to the substrate, which is essential to obtain good performance of the solar cells [49][50][51], to be assessed. A second inset to figure 3(a) contains a HRTEM image of the interface between the TiO 2 and CH 3 NH 3 PbI 3 . There is no gap between the materials, demonstrating good adhesion of the active layer to the TiO 2 . The PMMA layer shows perfect coverage of the semiconductor in figures 3(a) and (b), suggesting that it can act as an effective protective layer for the perovskite against atmospheric conditions that could result in premature degradation.
HRTEM analysis of the perovskite layer was carried out to study the crystal structure and quality of the sample in more detail. Figure 3(c) shows a HRTEM image of the CH 3 NH 3 PbI 3 layer in figure 3(a). The layer is formed from small crystallites, whose sizes range between 5 and 20 nm. Some of the crystallites are faceted, as shown in the two insets to figure 3(c). The presence of facets in the CH 3 NH 3 PbI 3 perovskite have been observed previously [52] and results from the minimization of total surface free energy according to thermodynamic theory [52][53]. FFTs of the HRTEM images were analyzed to identify the crystal structure of the CH 3 NH 3 PbI 3 perovskite. For comparative purposes, kinematic digital diffraction patterns (DDPs) of the three possible crystalline structures of CH 3 NH 3 PbI 3 (orthorhombic (Pnma), tetragonal (I4/mcm) and cubic (Pm3m)) were simulated using Eje Z software [38]. Distances and angles were measured from the experimental FFTs automatically in Digital Micrograph using the script diftool, which provides better precision than manual measurements. Figure 3(d) shows an experimental FFT obtained from a HRTEM image of the specimen, alongside a calculated DDP for the tetragonal crystalline structure. There is a good agreement between the patterns. The angles and distances measured from the FFT were introduced into the software, which compares the measured values with all of the possibilities for the three crystal structures. This analysis is consistent with a tetragonal crystal structure of the analyzed perovskite. It was carried out in several regions of the perovskite layer, as well as in the sample with a reduced precursor concentration, showing the same result. The results of the analysis are consistent with the PL measurements, which also suggest a tetragonal crystal structure. Although some previous studies have reported that the tetragonal structure of CH 3 NH 3 PbI 3 forms at room temperature and the cubic phase forms at higher temperature (>54°C) [54], others report that the two structures can coexist at room temperature [55]. The transition from a tetragonal to a cubic structure is accompanied by a slight distortion of PbI 6 octahedra about the c axis [54]. The crystal structure of the perovskites is important to understand their optical properties, as it has an effect, for example, on their band gap, with a difference of 0.03 eV reported between the band gaps of tetragonal and cubic CH 3 NH 3 PbI 3 [56].
Perovskite CsPbBr 3 NPs were also analyzed. Figure 4(a) shows a low magnification high-angle annular dark-field (HAADF) (S)TEM image of CsPbBr 3 NPs deposited on a Si substrate. The deposited perovskite particles are rectangular, with a size of ∼17±6 nm. Unlike for the CH 3 NH 3 PbI 3 layers, the fabrication process involved two steps. First, the perovskite material was synthesized in a colloidal solution and deposited onto the substrate. The synthetized material in solution was analyzed by TEM before deposition onto the substrate. CsPbBr 3 NPs synthesized at room temperature have rectangular shapes, and similar dimensions [20]. The sizes and shapes of the CsPbBr 3 particles are key to control the optical properties of the material [40], [57], [58], and can be tuned by modifying the solvent, ligand and reaction time [59]. We find that these characteristics of the CsPbBr 3 particles are not modified after their deposition onto a solid substrate, which is encouraging for their integration into solid state devices. In the film, the particles are deposited close to each other. However, some gaps can be observed between them. Subsequent particles are deposited on the initial ones, but do not seem to fill the small gaps. The PMMA layer shows perfect coverage of the perovskite particles (see figure 4(a)), just as for the CH 3 NH 3 PbI 3 sample.
The NPs have crystalline structures, as can be observed in the inset to figure 4(a). The crystal structure was studied using the procedure described above. In this case, the three possible structures of the CsPbBr 3 perovskite are cubic (Pm3m), tetragonal (P4/mbm) and orthorhombic (Pnma). Figure 4(b) shows an FFT of the image in the inset to figure 4(a), alongside a calculated DDP for a cubic structure. Good agreement is observed, suggesting that the crystal structure of the perovskite material analyzed is cubic and consistent with the results obtained from the PL study. This cubic structure has been reported to be advantageous for CsPbBr 3 perovskites for photovoltaic applications due to its optimal band gap of ∼2.4 eV [60] in NP form. Other low temperature synthesis methods lead to yellow and non-luminescent nanocrystals, as a result of poor crystallinity or the formation of an orthorhombic crystal structure [61]. However, it has been reported that the cubic structure suffers a transformation to the orthorhombic structure when it is exposed to ambient conditions [62]. Several options have been considered to stabilize the cubic structure. For example, purification with methyl acetate has proved to be a route to stabilize the cubic structure under ambient conditions for months [63]. The introduction of a layer of PMMA is also a potential solution, as it provides efficient coverage of the perovskite, as shown in the present work.
In order to study the compositional distribution in the perovskite layer, EDX analysis of the particles was carried out, as shown in figures 4(c) and (d). The EDX map shown in figure 4(c) confirms the expected uniform distribution of the elements. The slight increase in signal on the left side of the particle is thought to result from an increase in the thickness of the material in this region, as it is found for all of the elements. In figure 4(d), although the distribution of Br and Cs is uniform, the distribution of Pb is inhomogeneous. Pb 0 formation in halide perovskites under light illumination has been previously reported using XPS and XRD [64]. The EDX map shows a larger amount of Pb at the edges of the NP. An inset showing only the Pb distribution is included to show the enriched edges. Similar results were found in some other CsPbBr 3 NPs in this sample, although with a lower degree of segregation. Recent studies reported the degradation of hybrid and all-inorganic lead halide perovskites under electron beam irradiation in the TEM [33]. In particular, it was reported that CsPbBr 3 particles undergo a radiolysis process that results in the desorption of Br atoms and the reduction of Pb 2+ ions to Pb 0 [33], which diffuse and aggregate to form non-epitaxial Pb crystals [33]. The damage was found to be particularly pronounced at the corners and edges of the perovskite layer, as there is a lower diffusion barrier for Pb 0 in these regions [33]. Our results also point in this direction. The longer exposure times used for the acquisition of EDX maps in the electron microscope, in comparison to conventional imaging modes, results in the initial stages of material decomposition in some particles. As mentioned above, the damage of the CsPbBr 3 perovskites under the electron beam consists on the reduction of Pb +2 -Pb 0 and then the diffusion of Pb atoms to form Pb particles. The diffusion of Pb atoms is a procedure that needs time. When the exposure time is short, only a small amount of Pb +2 is reduced to Pb and there is not much time for the Pb diffusion. However, when the exposure time is longer as during the acquisition of EDX spectra, a larger amount of Pb +2 is reduced and diffuses to form Pb aggregates. The observed decomposition is more pronounced at the edges of the particles, as shown in figure 4(d). The edge of a material (as well as any other defected region) is a part of the material easier to be modified as the atoms do not have all their bonds completed, therefore initial Pb particles would be expected to be formed in this region, as shown previously by [33,65,66]. Also, the diffusion barrier is lower in this region, therefore some accumulation is expected to be found at the edge of the particles. Care is therefore required in the study of perovskite materials using (S)TEM techniques, in order to avoid degradation of the material and to obtain reliable results.

Conclusions
We have shown that specimen preparation using FIB milling instead of conventional techniques and the use of a low accelerating voltage of 80 kV are required for (S)TEM analysis of APbX 3 perovskites in order to avoid degradation of the material on both standard thin films and layers formed by NPs. Analysis of CH 3 NH 3 PbI 3 synthesized by a one-step method shows that the active layer grows irregularly on a Si substrate from the initial steps of the crystallization process, with good adherence to a TiO 2 layer. The crystal structure of the CH 3 NH 3 PbI 3 perovskite is shown to be tetragonal. A sharp absorption edge is localized at 750-760 nm, while the exciton absorption resonance is centered at 750 nm. CsPbBr 3 NPs deposited on a substrate are shown to have parallelepipedal shapes similar to those before the deposition process. Their crystal structure is cubic with a PL line centered at 517 nm with a FWHM of 21 nm, which is optimal for lightemitting diodes and displays with high color rendering. Despite the low voltage used for (S)TEM analysis, long irradiation times result in segregation of Pb to the edges of the NPs, which is the first step towards decomposition of the material. The analysis highlights the care that is needed for (S)TEM measurements of both bulk and NP perovskite thin films.