Unravelling Atomic Structure and Degradation Mechanisms of Organic-Inorganic Halide Perovskites by Cryo-EM

SUMMARY

Despite rapid progress of hybrid organic-inorganic halide perovskite solar cells, using transmission electron microscopy to study their atomic structures has not been possible because of their extreme sensitivity to electron beam irradiation and environmental exposure. Here, we develop cryogenic-electron microscopy (cryo-EM) protocols to preserve an extremely sensitive perovskite, methylammonium lead iodide (MAPbI₃) under various operating conditions for atomic-resolution imaging. We discover the precipitation of lead iodide nanoparticles on MAPbI₃ nanowire’s surface after short UV illumination and surface roughening after only 10 s exposure to air, while these effects remain undetected in conventional x-ray diffraction. We establish a definition for critical electron dose, and find this value for MAPbI₃ at cryogenic condition to be 12 e⁻/Ų at 1.49 Å spatial resolution. Our results highlight the importance of cryo-EM since traditional techniques cannot capture important nanoscale changes in morphology and structure that have important implications for perovskite solar cell stability and performance.

INTRODUCTION

Hybrid organic-inorganic halide perovskite solar cells (PSCs) have attracted tremendous research attention and quickly dominated photovoltaic (PV) research for a number of reasons. As a solution-processable material, hybrid perovskite (CH₃NH₃PbI₃, MAPbI₃) has the potential to deliver efficient, scalable and inexpensive solar energy.^1-4 The first solar cell employing hybrid perovskite was reported in 2009 with a power conversion efficiency (PCE) of 3.81%.⁵ In just one decade, PSCs now have reached a PCE > 24%, which is comparable to that of silicon-based solar cells that have enjoyed many decades of development.^2-4 Despite the rapid growth in performance efficiency, the commercial deployment of PSCs is hindered by their poor stability under practical operating conditions.^{3, 6-8}

As illustrated in Figure 1A, perovskite solar cells have been reported to decompose into precursor materials upon exposure to common environmental conditions such as moisture,⁶ local heating⁹ and ultraviolet (UV) illumination.^10-12 Other phenomena such as oxygen interpenetration and illumination-induced ion-migration^{13, 14} not only diminishes material stability by structural deformation, but also leads to anomalous hysteresis in current-voltage curves that are not yet fully understood. Despite significant research activity directed towards the application of PSCs, the atomic and nanoscale understanding of how degradation conditions change the crystal structure, morphology, and physical properties of hybrid perovskite remains elusive. Therefore, it is critical to develop methods for revealing the atomic structure of perovskite solar cells at various operating conditions to provide insight into addressing important issues with stability and pave the way for commercialization of perovskite optoelectronic devices.

Figure 1. Preserving and stabilizing hybrid perovskite using Cryo-EM.

(A) Upon exposure to e- beam, UV or moisture, hybrid perovskites decompose into the precursor materials (B) Standard TEM image of MAPbI₃ NWs. After exposure to electron dose ~ 500 e⁻/Ų, MAPbI₃ decomposes into PbI₂. (C) Low dose image of MAPbI3 NW inserted into TEM by standard TEM sample loading procedure at room temperature. NW surface roughening confirms the degradation during room temperature sample insertion. D) Pristine or UV/moisture exposed perovskite NWs were dropcast onto a quantifoil TEM grid in nitrogen glovebox. Then, the sample was plunge-frozen into liquid nitrogen inside glovebox. (E and G) TEM images of MAPbI₃ (E) and MAPbBr₃ (G) shows smooth surface on NWs. (F and H) Atomic-resolution TEM images resolving [PbI₆]⁴⁻ octahedral and MA⁺ molecule of MAPbI₃ (F) and MAPbBr₃ (H). Electron dose is ~ 12 e⁻/Ų in (C and F) and ~46 e⁻/Ų in (H)

Unfortunately, current understanding of degradation mechanisms is largely limited to ensemble studies using X-ray/neutron diffraction,^15-17 selected area electron diffraction (SAED)^{18, 19} and surface sensitive techniques.¹⁰ Although transmission electron microscopy (TEM) has been used to study the macroscopic morphology of perovskite solar cells, atomic-resolution imaging remains challenging due to its extreme sensitivity to electron beam damage.⁹ By imaging at ultralow electron dose (11 e⁻/Ų), Han et al. successfully acquired a high-resolution TEM (HRTEM) image of a bromide-based perovskite (MAPbBr₃) at room temperature.²⁰ However, MAPbBr₃ remains an inefficient perovskite solar cell material with limited application due to its large bandgap. In contrast, the iodide-based perovskite (MAPbI₃) is much more efficient yet cannot be imaged as readily with atomic-resolution since it is much more sensitive than MAPbBr₃, creating an extreme challenge for TEM imaging.

Recently, cryogenic-electron microscopy (cryo-EM) was shown to be a powerful tool beyond structural biology. In particular, reactive and electron beam-sensitive lithium battery materials have been successfully stabilized for TEM sample loading, imaging and spectroscopy.^21-26 We hypothesize such cryogenic condition may also reduce electron beam damage to hybrid perovskites such as MAPbI₃. Here, we establish a cryo-EM protocol to successfully reveal the atomic structure of MAPbI₃ and quantify its cryogenic critical dose to be 12 e⁻/Ų at 1.49 Å spatial resolution. We also measured the critical dose of MAPbBr₃ to be 46 e⁻/Ų, which is improved by 4 times compared to room temperature condition (only 11 e⁻/Ų) in previous study.²⁰ We measured that the electron dose tolerance of MAPbI₃ is much lower than MAPbBr₃ at the cryogenic condition. Furthermore, cryogenic conditions enable preservation of the pristine or degraded MAPbI₃ structure during various durations of UV light or moisture exposure. Therefore, the cryo-EM methodology developed in this study enables our investigation into the nanoscopic decomposition pathways of MAPbI₃ and enables correlating nanostructure with macroscopic device performance.

RESULTS AND DISCUSSION

Cryo-EM imaging for hybrid perovskite

Hybrid perovskites possess an ABX₃ structure, where A usually contains an organic cation (MA⁺ = methylammonium), B is a divalent metal cation (Pb²⁺), and X is a halide anion (I⁻, Br⁻ or Cl⁻) (Figure 1A). The vast majority of perovskite solar cells comprise a thin film geometry, where a polycrystalline thin film (> 300 nm) is formed with a wide range of processing methods, microstructures, and precursors. However, there is still a lack of mechanistic understanding on the control of film crystallinity and morphology. Additionally, optoelectronic properties have been shown to depend on perovskite composition²⁷ and grain boundaries.²⁸ Single crystals of pure MAPbX₃ have not only shown to exhibit among the lowest defect densities and highest charge carrier diffusion lengths,²⁹ but they also involve the simplest perovskite structure that is most commonly used in literature for high efficiency solar cells.³⁰ Due to the existence of heavy Pb atoms, a suitable perovskite TEM sample should not be thicker than 100nm. In addition, previous study has successfully demonstrated the fabrication of effective solar cells based on perovskite nanowires (NWs).³¹ As a result, single crystal MAPbI₃ NWs were chosen as a convenient model system to study the nanoscale degradation mechanisms of the solar cell material.

To exclude the effect from surface ligand,⁸ in this study, MAPbI₃ nanowires were synthesized on glass substrates using a previous reported ligand free method³² inside a nitrogen glovebox. The split of the (220) and (004) peaks in powder X-ray diffraction (XRD) (Figure S1) pattern indicates a pure tetragonal MAPbI₃ crystal structure (space group I4/mcm, a = 8.896 Å, c = 12.707 Å), rather than the high temperature cubic phase.³³ No residue PbI₂ precursor is observed. MAPbBr₃ NWs were synthesized by replacing MAI precursor with MABr. The XRD patterns show that the as-grown NWs are cubic phase MAPbBr₃ (space group Pm$\overline{3}$m) without impurities (NIHMS1583162-supplement-supporting.docx). However, as shown in Figure 1B, MAPbI₃ NWs quickly decompose into PbI₂ upon exposure to high electron dose (500 e⁻/Ų) during TEM imaging at room temperature due to extreme electron beam-sensitivity of the organic MA⁺ cation. Previous TEM studies on MAPbI₃ were usually conducted under low magnification using SAED with carefully tuned low electron dose rate of only 1 e⁻Å⁻²s⁻¹ to avoid potential beam damage at room temperature.^{18, 19} In addition, the short (< 5s) air exposure during standard TEM sample insertion will expose MAPbI₃ NWs to moisture. As shown in Figure 1C, after standard TEM sample insertion procedure, the MAPbI₃ NW was imaged with an electron dose of ~12 e⁻/Ų to avoid beam damage. A clear surface roughening is observed, which demonstrates the surface degradation during sample insertion.

To overcome the particular electron beam and moisture sensitivity of perovskite solar cell materials, we engineered a plunge-freezing procedure modified from cryo-EM methodologies used in structural biology³⁴ to preserve the hybrid perovskites in their pristine or operating state for high-resolution imaging (Figure 1D). First, the as-grown perovskite NWs were dropcast onto a Quantifoil TEM grid in an inert environment. Then, the sample was plunge-frozen into liquid nitrogen in the glovebox. Side reactions between hybrid perovskites and nitrogen or moisture is kinetically inhibited at cryogenic temperatures, which allows the plunge-freezing process to retain the original state of the hybrid perovskite NWs with the relevant structural information successfully preserved. To study the nanoscale degradation mechanism under independent UV light radiation or moisture exposure, we subjected samples to UV illumination (Figure S3) inside an Ar-filled glovebox or in ambient air (Relative Humidity ~40%) before plunge-freezing. The sample was then transferred into the TEM column (operating at 200kV) without any air exposure and kept at −175 °C for TEM imaging using a direct-detection electron-counting camera (DDEC).³⁵ With the high quantum detective efficiency in the counting mode of DDEC, images can be acquired under extremely low dose. In addition to electron beam damage, HRTEM of beam sensitive materials often suffers from beam-induced sample motion, which blurs the image.³⁶ With the high detective quantum efficiency of DDEC, we were able to acquire images by breaking down the total dose into a successive series of low-dose frames followed by subsequent frame alignment to minimize this motion³⁶ (see Methods).

Figure 1E is a typical Cryo-HRTEM image of MAPbI₃ NW taken at the magnification of 80,000 corresponding to a pixel size of 0.46 Å by 0.46 Å. In general, sample regions were exposed to an electron dose of ~12 e⁻/Ų to avoid beam damage. We noticed that higher electron dose may induce sample damage through the formation of methylamine or halogen bubbles³⁷ (Figure S4), which is not observed at room temperature. In contrast to beam-damaged MAPbI₃ NWs under standard room temperature TEM conditions, NWs imaged at cryogenic conditions have a much smoother surface. High-resolution images of MAPbI₃ NWs (Figure 1F) resolve individual [PbI₆]⁴⁻ octahedral and MA⁺ molecule column, showing that perovskite NWs are single crystalline. Image simulation confirms that the areas of bright contrast corresponds to MA molecule columns (Figure S5 and S6). A series of Fast Fourier transform (FFT) of the MAPbI₃ image (Figure 2A) shows structural information down to 1.49 Å. It has been reported that MAPbI₃ undergoes an equilibrium phase transition from tetragonal to orthorhombic at ~ 165K.^{17, 38} However, based on HRTEM images and the FFT (Figure 1F and 2A), this structural change is not observed likely due to the fast cooling of our plunge-freezing procedure, further demonstrating that the native state of our material (tetragonal phase) can be successfully preserved. In addition, HRTEM images of MAPbBr₃ and mixed halide MAPb(Br_0.5I_0.5)₃ NWs were taken at the same magnification (Figure 1G and Figure S7). We found that typically MAPbBr₃ and MAPb(Br_0.5I_0.5)₃ NWs are over 4 times more tolerant to electron beam exposure than MAPbI₃. Thus, to minimize the dose-rate for different perovskites and establish a standard for electron dose accumulation, we performed an electron dose tolerance test that dictates the ideal electron dose exposure for each material in our study.

Figure 2. Radiation damage measurement and quantification.

(A-C) Fast Fourier transform of MAPbI₃ HRTEM images with cumulative electron exposure 7.6 e⁻/Ų (A), 15.2 e⁻/Ų (B) and 22.8 e⁻/Ų (C). Yellow circle in (A) shows information as low as 1.49 Å is preserved. White circle represent an information of 2 Å. (D) Normalized intensity of diffraction spots with lattice spacing less than 2 Å vs cumulative electron exposure for MAPbI₃ (red) and MAPbBr₃ (blue). Electron dose that decreased normalized intensity to 30% from its maximum is marked with star as critical electron dose for HRTEM imaging. Dose rate of 3.8 e⁻ Å⁻²s⁻¹ is used for MAPbI₃, 25.6 e⁻ Å⁻²s⁻¹ is used for MAPbBr₃.

Radiation damage measurement and quantification

Electron beam damage has been investigated rigorously in biological cryo-EM as it is the fundamental factor that limits the resolution of frozen-hydrated specimens.³⁹ Here we extend the electron radiation damage evaluation methodology from biological cryo-EM to the beam-sensitive hybrid perovskite. With the aid of the electron counting mode of DDEC, the electron dose rate (e⁻Å⁻²s⁻¹) of certain illumination conditions can be characterized. A slightly longer exposure time, typically 3-4 s, was chosen to record a multiple-frame image stack. Figure 2A-C is a stack of successive FFT of MAPbI₃ NWs HRTEM images during a dose tolerance test (7.6, 15.2 and 22.8 e⁻/Ų). The white dashed circle represents an information transfer of 2 Å. When the number of frames taken and the cumulative electron exposure are increased, MAPbI₃ shows a partial loss of crystallinity (15.2 and 22.8 e⁻/Ų in Figure 2B, C, respectively).

Furthermore, the intensity of diffraction spots in FFT with lattice spacing less than 2 Å was extracted and normalized for quantitative analysis. Figure 2D shows the plot of normalized intensity as a function of cumulative electron exposure. The fluctuation of normalized intensity after long-term electron beam exposure is attributed to background noise. Based on this plot, we define the electron dose that decreased normalized intensity to 1/e from its maximum as the critical HRTEM imaging dose for perovskite, which is ~12 e⁻Å⁻² in the case of MAPbI₃ (marked by star in Figure 2D). It’s worth mentioning that under this definition, the information higher than 2 Å is still preserved as shown in Figure 2B, corresponding to 15.2 e⁻Å⁻² cumulative electron exposure. Similarly, a dose tolerance test was performed on MAPbBr₃ NWs (Figure S8). Indeed, as shown in Figure 2D, MAPbBr₃ has a higher electron beam tolerance compared to MAPbI₃. Following the same definition, we identify 46 e⁻Å⁻² as the critical dose for MAPbBr₃, which is 4 times higher than the electron dose reported previously.²⁰

Exposure to ultraviolet irradiation

Prolonged exposure of MAPbI₃ to UV light irradiation has been previously shown to decompose the material into PbI₂ via a free electron mechanism involving the TiO₂ conduction layer.¹¹ The stability of MAPbI₃ on nonconductive substrates like Al₂O₃ is believed to be much-improved based on XRD characterization.¹² However, the nanoscale degradation mechanism of MAPbI₃ remains elusive. Unfortunately, probing the nanoscale structure involved in this degradation process requires high spatial resolution techniques using electron microscopy. Previous studies have been inconclusive because it has not been possible to deconvolute the effect of the electron beam from the UV light degradation, both of which decompose the MAPbI₃ to PbI₂.

Using our cryo-EM methodology, we tested the intrinsic UV light stability of MAPbI₃ NWs without the addition of any charge transport layers (e.g. TiO₂). XRD spectra of MAPbI₃ NWs after broadband UVA light exposure in an argon environment for 0 min (pristine), 15 min, 30 min, and 60 min were collected (Figure 3A). These short exposure times did not form any detectable PbI₂ peaks (black PbI₂ reference curve in Figure 3A) in the XRD dataset. Furthermore, 60 min UV illuminated MAPbI₃ NWs was loaded on silicon (100) substrate for XRD fine scanning around PbI₂ peak positions. As shown in Figure S9, no PbI₂ peak was detected. Therefore, this data shows the MAPbI₃ NWs are stable after the short UV light illumination, especially because there is no TiO₂ layer. However, when the above UV-exposed MAPbI₃ NWs are investigated using the established cryo-EM methodologies and protocols described above, we discover an obvious change in nanowire surface morphology visible as dark contrast region (~5 nm) in the nanowire image that was not observed using conventional characterization methods (Figure 3B).

Figure 3. MAPbI₃ degradation under UV light radiation.

(A) XRD spectra of MaPbI after 0 min (pristine), 15 min, 30 min, and 60 min of UV illumination. No PbI₂ peaks (curve in black) are detected within an hour of UV illumination. (B-D) Cryo-EM images of MAPbI₃ NW after 15 min (B), 30 min (C), and 60 min (D) of UV illumination. Nanograins of crystalline PbI₂ are clearly resolved on the surface of the MAPbI₃ NW.

This morphology is markedly different from the pristine case when not exposed to UV light (Figure 1D). These dark contrast features continue to grow larger with increasing UV exposure (Figure 3C-D) within the time of our investigations. When examining the regions of dark contrast at a higher magnification, we can resolve atomic crystalline structures that correspond to the (110) lattice spacing of PbI₂ (PDF No. 07-0235). Our observation provides new insight to previous XRD studies that suggest PbI₂ would not be formed at short exposure times. In previous room temperature TEM experiments, this observation would be attributed to artifacts generated by the electron-beam induced degradation of the MAPbI₃ material. However, the cryo-EM methodology used in this study with a critical dose limited to 12 e⁻Å⁻² ensures that our observation remains an accurate reflection of the NWs’ states preserved during cryogenic freezing without the concern of artifacts generated by the imaging technique.

There are several important implications of this new discovery. Firstly, this degradation under short UV light exposure is unexpected and surprising because it could not be detected using previous techniques like XRD or UV-Vis spectroscopy. This may be because the nanograins of PbI₂ are too small (5 nm) and the number density is too low for XRD to detect. Thus, the apparent formation of PbI₂ could also influence the opto-electronic properties of the MAPbI₃, which was previously thought to be pristine. Second, the fact that this degradation occurs without the presence of TiO₂ suggests an entirely different degradation mechanism is at play when UV light is illuminated onto MAPbI₃ and implies that the material is intrinsically unstable. One initial hypothesis is that the UV light triggers a structural degradation in the sensitive organic-inorganic bonds, collapsing the framework structure and accelerating the decomposition back into the PbI₂ starting material. Finally, the apparent presence of PbI₂ decomposition product under such short UV exposure times could have an impact on the performance of individual devices and could potentially explain the broad range of device efficiencies within even the same batch. Additionally, there is a commonly reported effect of early-time light-induced degradation or “burn-in” reported during initial solar cell operation⁴⁰—the mechanism for which remains unknown—which may be related to the morphological changes observed by cryo-EM. This new finding has important implications for the performance of this important solar cell material and highlights the potential impact of new discoveries using nanoscale studies on degradation mechanisms enabled by our cryo-EM protocol.

Exposure to moisture environment

Moisture sensitivity of MAPbI₃ also limits its long-term stability and has been shown to greatly degrade device performance.⁷ Many theories on the degradation mechanism have been proposed based on the chemical reactions with water and oxygen, but are largely limited by the lack of nanoscale information. Conventional techniques like XRD that measures the average ensemble structure may not detect important nanoscale changes, as shown above for UV irradiation. Previous studies have shown that the MAPbI₃ XRD peak disappears after exposure to moisture on the timescale of days (~24 hours).⁸ However, an atomic level picture of the initial degradation process is lacking.

To reveal the nanoscale degradation process of MAPbI₃ NWs, we exposed MAPbI₃ NWs to air with 40% relative humidity at room temperature for short durations before plunge-freezing into liquid nitrogen to preserve its degraded state. Optically, the color change from dark brown to yellow indicates full decomposition of the MAPbI₃ after 2 days of moisture exposure (Figure 4A). At shorter exposure times of 10s, 45 min and 2 hours, no change can be detected optically and PbI₂ peaks indicative of degradation remain absent in XRD spectra (Figure S10), which would suggest the MAPbI₃ NWs remain stable. In contrast, high-resolution cryo-EM images uncover the initial nanoscale degradation process that cannot be detected using conventional techniques. Surprisingly, even with only 10 s of moisture exposure, a clear roughening of the surface is observed (Figure 4B, C). This extreme sensitivity in air is likely due to the lack of surface-protecting ligands during the synthesis of these perovskite NWs, which means that a standard TEM procedure could cause artifacts during perovskite sample loading.

Figure 4. MAPbI₃ degradation under moisture exposure.

(A) Optical images of MAPbI₃ NW films taken at various exposure durations in air with 40% relative humidity at room temperature. Color change is visible after 2 days. (B-G) Cryo-EM images of MAPbI₃ NWs after 10s (B and C) 45 min (D and E) and 2 hours (F and G) of moisture exposure. (H) Schematic of proposed nanoscale degradation mechanism of MAPbI₃ in moisture environment: a thin amorphous film (~5 nm) initially forms on the surface of the MAPbI₃ that facilitates further decomposition into PbI₂ and other components (e.g. HI and CH₃NH₂).

After 45 min of moisture exposure, a thin amorphous layer (~5 nm) appears on the surface of the MAPbI₃ nanowire (Figure 4D, E). The amorphous material could be a hydrated compound that forms when exposed to water vapor.⁷ The FFT of the MAPbI₃ (Figure 4D inset) indicates that the crystallinity of the nanowire as a whole has not been damaged, yet the surface is clearly degraded. This could have important implications on charge-transport phenomenon at the surface of MAPbI₃ and its interface with charge carrier conduction layers. Light harvesting would also change with the growth of this amorphous film upon exposure to moisture. After 2 hours of moisture exposure, cryo-EM images show that the MAPbI₃ NW surface has noticeably become rougher with the presence of whisker-like structures sticking out (~30 nm) (Figure 4F, G). High-resolution images reveal these nanoscale whiskers contain nanograins of PbI₂ (~5 nm) dispersed in the amorphous layer observed as in Figure 4D, indicating that the degradation process has progressed after only 2 hours of moisture exposure.

The high-resolution studies of moisture-induced MAPbI₃ degradation enabled by cryo-EM suggest a nanoscale degradation pathway (Figure 4H). Initially, an amorphous layer likely consisting of a hydrated compound coats the MAPbI₃ surface. This thin layer then facilitates decomposition of the MAPbI₃ into gaseous components (e.g. HI and CH₃NH₂) and nanograins of PbI₂, further roughening the surface of the remaining MAPbI₃ material. Interestingly, this degradation proceeds under much milder conditions (no UV irradiation, room temperature) than previous studies, which could have important implications on materials processing protocols and requirements for device sealing to minimize exposure to moisture. This result could explain why some of the most stable reported solar cells use procedures that eliminate moisture exposure by depositing in-situ barrier layers⁴¹ or encapsulating in inert environments.⁴²

Outlook and Conclusion

The protocols established in this study introduces a specimen preparative method for sensitive materials systems like hybrid perovskite solar cells under experimentally defined environmental conditions including UV illumination and moisture exposure. By engineering the freezing process to preserve the authentic nanostructures at nanoscale and closely monitoring the cumulative electron exposure to minimize its radiation damage effects, we successfully retrieve the structure signatures of MAPbI₃ solar cell materials at atomic resolution level. Single crystal MAPbI₃ NWs were used as a model system to develop a fundamental understanding of degradation without the confounding effects of grain boundaries and to maximize interface area. The next step characterizing perovskite devices with cryo-EM will leverage a process for cross-sectioning full perovskite devices using a cryo-FIB process that has been widely applied in biology field. This will enable the observation of device-specific degradation modes that result from operational conditions, such as interface chemistry, halide segregation in solar cells and halide exchange in LEDs. The developed protocol will open up future studies of more complex perovskite materials (i.e. 2D Ruddlesden-Popper perovskites) for optoelectronic applications and other electron-beam sensitive materials with unprecedented understanding of their atomic scale structures and chemistries.

EXPERIMENTAL PROCEDURES

Synthesis of Hybrid Perovskite Nanostructures

First, lead iodide (PbI₂) was spin-coated onto glass slide (1cm x 1.5cm) at 3000 rpm for 30s from a solution of 400 mg/mL PbI₂ (99.99% TCI) in dimethylformamide (DMF). Then, inside a nitrogen glovebox, the PbI₂ coated glass slide was immersed into 2 mL CH₃NH₃I (MAI), CH₃NH₃Br (MABr) or MAI_0.5Br_0.5 precursor solution with PbI₂ side facing up, to synthesize MAPbI₃, MAPbBr₃ and MAPb(I_0.5Br_0.5)₃ NWs respectively. The precursor solution was prepared by dissolving MAI, MABr or MAI_0.5Br_0.5 isopropanol (IPA) with a concentration of 20 mg/mL. Before spin-coating, the substrates were sequentially cleaned in ultrasonic baths of Extran 300 detergent (diluted 1:10 in DI water), DI water, acetone, and IPA followed by 15 minutes of UV-ozone treatment (Jelight UVO-Cleaner Model 42). PbI₂ coated glass was heated at 100 °C inside a nitrogen glovebox to remove the remnant solvent and water before placed into the precursor solution. After 12 hours, the glass substrate was removed and rinsed with 1 mL anhydrous isopropanol (SigmaAldrich) to remove any leftover salt on the surface, dried under a stream of nitrogen flow, and then dried on a hot plate at 100 °C for another 30 s.

X-ray Diffraction

As-synthesized or UV illuminated perovskite nanowires on glass was sealed with Kapton tape to avoid degradation from moisture during X-ray diffraction measurement. Then θ–2θ measurements were performed using a Panalytical X’Pert Pro Diffractometer (Copper anode, Kα1 =1.54060 Å, Kα2 = 1.54443 Å, Kα2 / Kα1 ratio = 0.50).

Electron Microscopy and Data Processing

Room temperature TEM characterizations were carried out using a FEI Titan 80-300 environmental TEM operated at 200 kV. All Cryo-EM experiments were performed on a Thermo Fisher Tecnai F20 transmission electron microscope operated at 200 kV. Cryo-TEM images were acquired by a Gatan K2 direct-detection camera in the electron-counting mode with the Dose Fractionation function. Considering the radiation sensitivity of the material, continuous exposure to adjust defocus is impractical. Multiple short time exposure single frame shots were taken to estimate the defocus and make it as close as possible to Scherzer defocus. Following the established data processing procedures in latest Cryo-EM makes this method more generalizable and accessible. Dose fraction frames were motion-corrected by MotionCor2⁴³ to correct beam-induced movements (Figure S11). The conditions of defocus and astigmatism are determined by CTFFIND4⁴⁴ and GCTF⁴⁵ followed by a manual examination. The motion-corrected images were then phase flipped by a set of MRC programs⁴⁶ that used to handle 2D protein crystal Cryo-TEM images in traditional 2D electron crystallography. All above procedures can be done in a modern software package FOCUS⁴⁷ with a user-friendly GUI or separately. The amplitude correction is not included in our CTF correction to avoid artificial effects. Finally, the CTF-corrected images were denoised by a Wiener filter and an Average Background Subtraction Filter (ABSF) filter⁴⁸ that commonly used in HRTEM to give better results. Both filters were applied to every image to avoid potential artifacts introduced by a specific filter. EMAN2⁴⁹ was utilized to sum up subsequent subsets of frames according to their exposure time from motion-corrected image stack, during radiation damage measurement and quantification. In our study, we summed every 5 frames per subset in the image stack. Intensity of diffraction spots in certain resolution shell in Fourier power spectrum was extracted and measured. The measured Fourier intensities were normalized by Minmax normalization. Minmax normalization is a normalization strategy which linearly transforms x to y= (x-min)/(max-min), where min and max are the minimum and maximum values in X, where X is the set of observed values of x.