Enhancing stability and efficiency of perovskite solar cells with crosslinkable silane-functionalized and doped fullerene

The instability of hybrid perovskite materials due to water and moisture arises as one major challenge to be addressed before any practical application of the demonstrated high efficiency perovskite solar cells. Here we report a facile strategy that can simultaneously enhance the stability and efficiency of p–i–n planar heterojunction-structure perovskite devices. Crosslinkable silane molecules with hydrophobic functional groups are bonded onto fullerene to make the fullerene layer highly water-resistant. Methylammonium iodide is introduced in the fullerene layer for n-doping via anion-induced electron transfer, resulting in dramatically increased conductivity over 100-fold. With crosslinkable silane-functionalized and doped fullerene electron transport layer, the perovskite devices deliver an efficiency of 19.5% with a high fill factor of 80.6%. A crosslinked silane-modified fullerene layer also enhances the water and moisture stability of the non-sealed perovskite devices by retaining nearly 90% of their original efficiencies after 30 days' exposure in an ambient environment.

O rganic-inorganic halide perovskite materials as light harvesters for new-generation photovoltaic application have been attracting tremendous attention in both scientific and industrial communities in the past few years [1][2][3][4][5][6][7][8][9] . The certified power conversion efficiency has skyrocketed from 3.8 to 22.1% (http://www.nrel.gov/ncpv/images/efficiency_chart. jpg.), owing to the material's intriguing optoelectronic properties such as its high absorption coefficient 10 , its high charge carrier mobility and lifetime 11,12 , and its long carrier diffusion length [13][14][15] . A variety of perovskite photovoltaic device architectures have been designed ranging from mesoscopic to planar structures with n-i-p or p-i-n layouts 6,10,[16][17][18] . Although the highest efficiency obtained in mesoscopic type devices is already higher than the commercial CIGS and CdTe thin film solar cells, the intrinsic instability of perovskite devices owing to moisture and water hampers their practical application in ambient conditions [19][20][21] . Recently, a number of endeavours have been directed at improving the long-term stability of perovskite devices. One strategy is to develop new two-dimensional layered perovskite materials, such as (C 6 H 5 (CH 2 ) 2 NH 3 ) 2 (CH 3 NH 3 ) 2 [Pb 3 I 10 ]. Though this layered perovskite film showed enhanced moisture resistivity, the enlarged bandgap and exciton binding energy resulted in a low power conversion efficiency (PCE) of only 4.4% (ref. 22). Applying a water-resisting layer on the surface of perovskite film was demonstrated as another effective approach to enhance the moisture resistance of regular perovskite devices. Alkylphosphonic acid o-ammonium, polyethylene glycol, tertiary and quaternary hydrophobic alkyl ammonium cations have been assembled in the bulk or on the surface of perovskite film as a moisture-resisting layer, yielding device efficiency in the range of 15.0-16.0% (refs 23-25). Though the moisture stability of perovskite film indeed was improved in those designs, the widely used spiro-OMeTAD hole transport layers (HTLs) still limit the device stability due to the moisture affinity of Li salt dopants [19][20][21] . Thus there is growing interest in a p-i-n planar heterojunction structure, which generally contains fullerene (typically (6,6]-phenyl-C 61 -butyric acid methyl ester (PCBM)) as the electron transport/collection layers (ETLs) because of several advantages they have such as photocurrent-hysteresis-free devices, low-temperature processing and high device efficiency 17,18,26,27 . Unfortunately, the relatively thin fullerene layer compared with the thick spiro-OMeTAD layer is not sufficient to effectively protect the perovskite layer from moisture damage in a humid environment. To tackle this issue, You et al. 28 replaced fullerene with thicker ZnO ETL in p-i-n perovskite devices. Enhanced device stability was demonstrated, but the average device performance of B15.0% is still far inferior to their fullerene counterpart 17,27 . More recently, an inorganic Ti(Nb)O x buffer layer has been introduced on top of the PCBM layer to improve the device stability 29 . The judicious control of the heavily doped inorganic layer may induce much complexity for practical production. Therefore, it is still essential and urgent to develop a facile route for enhancing the moisture resistance of perovskite devices without sacrificing photovoltaic performance.
In this manuscript, we report a water-resistant crosslinkable silane-functionalized fullerene ETL to improve the moisture stability of p-i-n planar perovskite solar cells. We also introduce doping to the crosslinked silane-modified fullerene layer so that its conductivity is not compromised by the crosslinking process. The combination of crosslinking and doping has resulted in both high efficiency and stable perovskite solar cells in an ambient environment without resorting to encapsulation techniques.

Results
Formation of water-resistant fullerene layer. The concept of crosslinking fullerene is illustrated in Fig. 1a. The planar heterojunction p-i-n perovskite solar cells studied in this work have a structure of transparent conductive electrodes (TCE)/hole transport layer (HTL)/perovskite/ ETL/top electrode. Here C 60substituted benzoic acid self-assembled monolayer (C 60 -SAM) whose chemical structure is shown in the Fig. 1a was applied to form hydrogen bonding with the crosslinking agent. Trichloro(3,3,3-trifluoropropyl)silane, which readily hydrolyses to form three hydroxyl groups (-OH), was selected as the crosslinking agent. The carboxyl group (-COOH) of the C 60 -SAM material easily formed a strong hydrogen bond with one of the hydroxyl groups on silane, and the silicon-oxygen (Si-O) bonds generated from the silane-coupling reaction crosslinked the C 60 -SAM and silane molecules together. In addition, the trifluoromethyl groups (-CF 3 ) from the silane materials made the crosslinked C 60 -SAM layer more hydrophobic, which prevented the moisture and water from penetrating into the perovskite layer. Therefore, the crosslinked C 60 -SAM layer acted not only as the electron transport and passivation layer, but also as a waterresistant layer to protect the perovskite film against damage by moisture.
To fabricate the perovskite devices, the CH 3 NH 3 PbI 3 (MAPbI 3 ) perovskite films were prepared by the two-step interdiffusion method 30 , where lead iodide (PbI 2 ) and methylammonium iodide (CH 3 NH 3 I, MAI) were sequentially spun onto the poly(bis(4-phenyl)(2,4,6-trimethylphenyl)amine) (PTAA)-covered indium tin oxide (ITO) substrates, followed by a solvent annealing process 10 . A smooth and compact perovskite film was obtained with an average grain size much larger than the thickness of the films 27 . C 60 -SAM was then spun on top of the perovskite films and thermally annealed at 100°C for 60 min. To crosslink C 60 -SAM, 5-8 ml of trichloro (3,3,3-trifluoropropyl) silane was added at the edge of the 5 inch diameter petri dishes during the thermal annealing process. The silane vapour was expected to penetrate into and crosslink the C 60 -SAM layer. The devices were completed by thermal evaporation of a buffer layer and a top electrode. More detailed information on film and device fabrication can be found in the experiment section. Figure 1b shows a typical cross-section scanning electron microscopy (SEM) image of a device. To find out the accurate thickness of the crosslinked C 60 -SAM layer, high-resolution transmission electron microscopy of the cross-sectional area of the film was carried out. The thickness of the optimized crosslinked C 60 -SAM layer was found to be roughly 20 nm.
Fourier transform infrared spectroscopy (FTIR) transmission spectra of C 60 -SAM before and after silane modification were collected to verify the formation of crosslinked C 60 -SAM. C 60 -SAM films on ITO substrates with and without silane modification were prepared by following the same procedure as that for device fabrication but with larger thickness, and the collected powder was subjected to FTIR analysis. As shown in Fig. 2a, several new bands clearly appeared in the corresponding FTIR spectrum after crosslinking with the silane agent. The strong vibration bands at 1,262 and 1,209 cm À 1 can be ascribed to the -CF 3 groups added to C 60 -SAM. As illustrated in Fig. 1, we speculate that one free -OH group of silane molecule forms a hydrogen bond with the -COOH group of C 60 -SAM, and the other two -OH groups are involved in the silane coupling reaction to form a Si-O-Si bond. The strong and wide vibration peaks (1,126-1,002 cm À 1 ) shown in Fig. 2a confirm the formation of Si-O-Si bonds 31,32 . Though the hydrogen bonds of Si-OH Á Á Á -COOH do not directly correspond to a vibration peak in the FTIR spectrum, they should affect the Si-OH and C ¼ O chemical bonds by increasing their vibration frequency. Previous studies 33 indicate that the isolated Si-OH groups show a sharp band at 3,690 cm À 1 . When the silane agent crosslinked with the C 60 -SAM layer via hydrogen bonding, the C ¼ O vibration band moved from 1,686 cm À 1 to a higher wavenumber of 1,705 cm À 1 . In addition, the appearance of a broad band at frequency region of 3,400-3,200 cm À 1 indicates that the Si-OH groups are hydrogen-bonded 33 . These results demonstrate well that the C 60 -SAM layer was successfully crosslinked with the silane coupling agent through hydrogen bonding.
Improved morphology and crystallinity of fullerene film. SEM was used to examine the surface morphology of the C 60 -SAM film after silane functionalization. As shown in Fig. 2b, some aggregates can be seen on the non-crosslinked C 60 -SAM film, while the silane-crosslinked C 60 -SAM film seems to be smoother, compact and uniform (Fig. 2c). The aggregates found on the C 60 -SAM film may be caused by the intermolecular hydrogen bonding between -COOH groups during baking. When silane was introduced during thermal annealing of C 60 -SAM, most of the hydrogen bonds were preferably formed between -OH groups of the silane molecule and -COOH groups of the C 60 -SAM due to the steric effect 34,35 , though the intermolecular interaction among neighbouring and closely packed C 60 -SAM molecules may still have formed upon film formation. We hypothesize that the hydrogen bonding (-OH Á Á Á -COOH) aligns the C 60 -SAM molecules in order, driven by the silane crosslinking process, instead of randomly forming dimers, and it extends the intermolecular interaction for considerable distances, which results in pseudo-polymer type structures. Thus a more uniform and compact C 60 -SAM film is produced, which was revealed by the SEM study shown in Fig. 2c. To further verify that the ordering of C 60 -SAM can be induced by the crosslinking functionalization, we performed X-ray diffraction measurements of the C 60 -SAM films on glass substrates with and without crosslinking treatment. Figure 2d clearly shows three diffraction peaks including a sharp diffraction peak at 12.4°in the X-ray diffraction pattern of the crosslinked C 60 -SAM film, whereas there is no obvious diffraction peak in the non-crosslinked C 60 -SAM film. These three peaks can be indexed as the (111), (220) and (311) diffraction peaks from a face-centered cubic (fcc) lattice [36][37][38] , indicating that C 60 -SAM have the same fcc structure as C 60 . It is worth noting that among all possible crystal structures of C 60 , the fcc structure has the best electron mobility 39 . The better crystallization of crosslinked C 60 -SAM supports our hypothesis of good alignment of C 60 -SAM by crosslinking treatment, and explains the improved device open circuit voltage (V OC ) as shown below 17 .
To evaluate the water-resistivity change of the C 60 -SAMcovered perovskite films by crosslinking, we first tested the contact angles of the C 60 -SAM layers deposited on perovskite films with and without crosslinking. The contact angle of the C 60 -SAM covered perovskite film without crosslinking was as low as 46°. The colour of the water droplet changed to yellow after 1-2 s as shown in Fig. 2e, indicating a quick decomposition of MAPbI 3 to PbI 2 . In contrast, the water resistance of the crosslinked C 60 -SAM film with hydrophobic -CF 3 functional groups was significantly improved, showing a much larger contact angle of 86° (Fig. 2f) and delayed film colour change up to 120 seconds upon contacting water. These results confirmed the successful functionalization of the C 60 -SAM film by the introduction of the silane agent, which significantly enhanced water-resistivity.
Doping and crosslinking effect. We characterized the performance of the devices before and after silane-functionalization of the C 60 -SAM layer. It is noted that a relatively thick fullerene layer is needed to enhance the water resistivity, while the larger thickness and cross-linking process inevitably increases the device contact resistance at the cathode side. To improve the electron extraction capability of the crosslinked C 60 -SAM ETL, a small amount of MAI was blended in the C 60 -SAM as n-dopant via the  ARTICLE electron transfer between the anions of the n-dopants and C 60 -SAM 40,41 . Lewis base anions (I À from MAI dopants) act as electron donors. Initially the electron density is localized on iodide when two counter ions are effectively separated from each other. As the iodide approaches fullerene upon solvent drying, the energy decreases and the electron density starts to redistribute over iodide to fullerene, which enhances the free electron density of the C 60 -SAM matrix to achieve higher conductivity 40,42 . Figure 3a shows the impact of crosslinking and doping of C 60 -SAM layer on the photocurrent density-voltage (J-V) curves of the corresponding perovskite solar cells. A typical perovskite device with a non-crosslinked C 60 -SAM layer had a short-circuit current density (J SC ) of 18.6 mA cm À 2 , a V OC of 0.75 V and a fill factor (FF) of 43% yielding a very low PCE of 6.0%, which may be explained by the bad morphology as well as the bad electron transport property of the C 60 -SAM layer. When the C 60 -SAM layer was crosslinked, the device V OC and FF increased dramatically to 1.05 V and 73.9%, respectively. The device using crosslinked C 60 -SAM delivered an efficiency of 17.0%, which is approaching that of the devices made with conventional PCBM 17,27,43 . MAI as n-dopant was then introduced into the crosslinked C 60 -SAM layer, and the changed conductivity was found to have a significant impact on the device performance, particularly FF. After doping the crosslinked C 60 -SAM with 10 wt.% MAI, the device showed a decent FF of 80.0%, resulting in a significantly enhanced PCE of 19.3%. Further increasing doping ratio to 20 wt.% deteriorated the device performance. All the photovoltaic parameters of the perovskite devices based on non-crosslinked C 60 -SAM, crosslinked C 60 -SAM and crosslinked C 60 -SAM with different MAI doping levels are summarized in Table 1. As shown in Fig. 3b, the optimized device based on doped crosslinked C 60 -SAM exhibited a J SC of 22.6 mA cm À 2 , a V OC of 1.07 V and a FF of 80.6%, yielding a PCE of 19.5%. No obvious photocurrent hysteresis was observed in the device made with doped and crosslinked C 60 -SAM by changing the sweeping direction and sweeping rate (Fig. 3b), indicating the effective fullerene passivation 26 . The integrated J SC from external quantum efficiency (EQE) spectrum shown in Fig. 3c reached 22.2 mA cm À 2 , which is in good agreement with that from J-V measurement. The steady-state photocurrent and efficiency measured at the maximum power point (0.93 V) are presented in Fig. 3d, which confirms the device performance parameters extracted from the J-V curve and verifies the absence of photocurrent hysteresis in our devices. Prolonged steady-state performance up to 2,000 s can be found in Supplementary Fig. 1. The statistics of FF and PCE distributions shown in Fig. 3e,f demonstrate the reliability and repeatability of the FF and PCE enhancement obtained by the crosslinking and doping effect.

Discussion
The significant enhancement of FF to over 80% can be explained by the reduced contact (or series) resistance (R S ) by doping the C 60 -SAM layer. The series resistance of devices with different ETLs was derived by fitting the slope of J-V curves. As shown in Table 1, the device R S was reduced from 8.19 O cm 2 to 3.72 O cm 2 simply by crosslinking the C 60 -SAM. R S was further reduced to 2.08 O cm 2 via doping 10 wt.% of MAI into crosslinked C 60 -SAM.
To verify the increased conductivity through crosslinking and doping, conductivity of the C 60 -SAM films was measured with co-planar films made on glass. The obtained current-voltage curves were presented in Fig. 3g. The non-crosslinked C 60 -SAM film showed a quite low conductivity of 8.4 Â 10 -5 S cm À 1 . After silane treatment, the conductivity of crosslinked C 60 -SAM increased significantly to 1.9 Â 10 À 3 S cm À 1 . By introducing 10 wt.% of MAI for n-doping, the conductivity was further boosted to 1.1 Â 10 À 2 S cm À 1 , which is over 100 times higher than that of non-crosslinked C 60 -SAM. Thus, the conductivity variation follows the same trend with that derived from J-V curves, though a different increasing factor was observed here,  which can be explained by the different transport processes measured. The conductivity enhancement by crosslinking C 60 -SAM can be explained by the significantly improved film morphology. A more compact packing/connection of C 60 -SAM by crosslinking is expected to reduce activation energy (E a ) for the hopping of electrons between the C 60 -SAM molecules [44][45][46] . In addition, the extended intermolecular overlap through hydrogen bonding in the crosslinked C 60 -SAM layer increased the electron density of the aromatic system of the C 60 -SAM molecules 44 , which produced a substantial decrease in E a . It should be noted that the doping of C 60 -SAM does not necessarily need to be coupled with the crosslinking to enhance the device performance.
In devices with PCBM rather than C 60 -SAM as ETL, we also observed an enhanced efficiency to 19.1% by simply doping PCBM with 5 wt.% of MAI, as shown in Fig. 3h. Detailed photovoltaic parameters can be found in Supplementary Table 1. This doping also explained the recently observed very large FF devices where excess amount of MAI was used in fabricating the perovskite films which simultaneously doped PCBM 47,48 .
To understand how doping and crosslinking of C 60 -SAM boosts the perovskite photovoltaic performance, the recombination dynamics process was examined by impedance spectroscopy (IS) measurement, which was widely utilized in analysing many photovoltaic cells, such as dye-sensitized solar cells 49,50 , organic solar cells 51 and perovskite solar cells 26,52,53 . The measured impedance spectra of devices with non-crosslinked C 60 -SAM and doped crosslinked C 60 -SAM present different characteristics. As shown in Fig. 4a, the impedance spectrum of the devices with doped crosslinked C 60 -SAM have one semicircle, indicating only one recombination channel dominates in these devices. It could be assigned to the traps-assisted charge recombination at the film surface or grain boundaries because the perovskite thin films have excellent crystallinity and low density of traps 54 . The equivalent circuit used for these devices is shown in Fig. 4b. In sharp contrast, the impedance spectrum of the devices with noncrosslinked C 60 -SAM show two semicircles with different time constants (Fig. 4c), indicating that two recombination channels exist in these devices. Besides the surface or grain boundary recombination, the large density of traps in C 60 -SAM is most likely the origin of the additional recombination process. As a result, an additional resistor-capacitor (RC) circuit was added into the equivalent circuit (Fig. 4d) to separate the individual contribution of the two recombination channels. Figure 4e shows the recombination lifetime of different recombination channels extracted from the IS modelling. The fitted values from impedance spectra of perovskite devices with 10D-CLCS and N-CLCS can be found in Supplementary Fig. 2. For the devices with non-crosslinked C 60 -SAM, the shorter time constant (t f ) was assigned to the recombination process in the C 60 -SAM, because it is almost invariant to the applied bias and mainly determined by the defects in C 60 -SAM. The longer time constant (t s ) was assigned to the recombination process in perovskite. Apparently, the device with doped and crosslinked C 60 -SAM has much longer recombination lifetime, proving a better passivation effect by crosslinking C 60 -SAM.
The main motivation to crosslink the C 60 -SAM layer is to enhance the water resistance of the perovskite devices, though we did see an enhancement of device performance simultaneously. We first tested the stability of these devices under extreme conditions by dropping water on top of the devices for a certain time. Figure 5a,b shows the photograph (extracted from Supplementary Movie 1) of devices based on conventional PCBM and crosslinked C 60 -SAM ETLs after exposure to water droplets for 4 min. It is noted that the perovskite device with conventional PCBM completely decomposed in 30 s after dropping water on it, while the device with crosslinked C 60 -SAM was much more stable and no colour change was observed over 4 min, indicating no decomposition of perovskite to PbI 2 occurred. More details can be found in the Supplementary Movie 1. Thereafter, we monitored the stability of perovskite devices using both conventional PCBM and self-developed doped crosslinked C 60 -SAM as a function of storage time in air. All the devices were stored without encapsulation in an ambient environment. The original J-V curves showing the degradation can be found in Supplementary Fig. 3 for both types of perovskite devices stored at ambient environment for various days. As shown in Fig. 5c, in terms of the PCE of typical devices based on the conventional PCBM, the majority of initial performance was lost dramatically in 1 week. While the typical perovskite devices using doped crosslinked C 60 -SAM demonstrated encouragingly much better stability and maintained almost 90% of the original device PCE after 30 days. The variation of key photovoltaic parameters versus storage time is summarized in Fig. 5d-f. We also carried out a stability study of the unsealed perovskite devices with both PCBM and crosslinked C 60 -SAM ETL under continuous illumination (20 mW cm À 2 ) in highly humid air (50-75% relative humidity). The devices were removed at an interval of 24 h to record the photovoltaic performance. The variation of key photovoltaic parameters versus illumination time is summarized in Supplementary Fig. 4. The original J-V curves showing the degradation can be found in Supplementary Fig. 5 for both types of perovskite devices under continuous light illumination in humid air. As shown in the Supplementary Fig. 4, a typical device based on the conventional PCBM lost the majority of initial performance in 36 h. In striking contrast, the typical perovskite devices using crosslinked and doped C 60 -SAM demonstrated encouragingly much better stability and still maintained roughly 70% of the original device PCE even after 168 h under light soaking. This represents a significant improvement of device performance compared with what was reported. Previous study on the stability of unsealed perovskite devices under continuous light illumination in humid air showed that the photovoltaic performance usually degraded quickly to less than 70% of their original efficiency within 48 h (refs 25,55,56). Such exceptional ambient stability can be attributed to the crosslinkable silane functionalized and doped fullerene, which forms an internal stable hydrophobic encapsulating layer while maintaining excellent electron transporting properties as a result of n-doping.
In summary, we reported a specific crosslinked C 60 -SAM ETL by bonding crosslinkable silane coupling agent, which have hydrophobic functional groups, onto C 60 -SAM. Such ETL successfully blocks any accessible pathways of water molecule permeation and thus protect the perovskite films underneath from moisture-caused damage. The results show that the crosslinking process can also improve the transport properties of ETL. In addition, MAI was demonstrated as an effective n-dopant for C 60 -SAM to further enhance its conductivity. The perovskite devices fabricated with doped and crosslinked C 60 -SAM exhibit much longer recombination lifetime in    Subsequently, all the films were put on a hotplate and covered by a glass petri dish. For the devices with crosslinked C 60 -SAM, 5-8 ml of trichloro(3,3,3-trifluoropropyl)silane (Sigma-Aldrich) was added at the edge of the petri dish during the thermal annealing process (100°C for 60 min). The silane vapour was expected to be able to penetrate into the C 60 -SAM layer and crosslink with C 60 -SAM. C 60 (20 nm) and bathocuproine (8 nm) were then thermally evaporated as the buffer layers. The devices were completed by thermal evaporation of Cu (80 nm) as the top electrode. The device working area was 8 mm 2 , as defined by the overlap of the ITO substrate and the Cu cathode. A small amount (5-20 wt.%) of MAI (pre-dissolved in 2-propanol) was blended in the C 60 -SAM solution before spin-coating to dope it.
Film and device characterization. The photocurrent density-voltage (J-V) curves of the devices were measured under AM 1.5 G irradiation (100 mW cm À 2 ), which was produced by a xenon-lamp based solar simulator (Oriel 67005, 150 W Solar Simulator). The light intensity was calibrated by a Schott visible-colour glass-filtered (KG5 colour-filtered) Si diode (Hamamatsu S1133) before photocurrent measurement. Keithley 2400 Source-Meter was used for recording the J-V measurements. The J-V testing was performed with forward scan direction (if not specified in the manuscript) at 0.1 V s À 1 and sweep delay time was 50 ms. No preconditioning protocol was used before the characterization. A shadow mask with an aperture area of 6.44 mm 2 was used for the photocurrent calibration (Supplementary Fig. 6 and Supplementary Table 2). The steady-state PCE was measured by recording the photocurrent at a bias voltage of 0.93 V. We calibrated the measured photocurrent density with integrated J sc from EQE measurement. Calibration with Oriel Si detector was carefully carried out before EQE measurement. The EQE was obtained using a Newport QE measurement kit by focusing a monochromatic beam of light onto the devices. The X-ray diffraction patterns of C 60 -SAM films were obtained by a Rigaku D/Max-B X-ray diffractometer in the Bragg-Brentano parafocusing geometry. A conventional cobalt target X-ray tube equipped in the diffracted-beam monochromator was set to 40 kV and 30 mA. Impedance spectroscopy of the device was measured by the E4980A Precision LCR Meter from Agilent with homemade software under 1 sun illumination at room temperature. The resistivity measurements of various C 60 -SAM films were performed using a four-point probe setup as described in the previous work 43 . C 60 -SAM films were deposited on normal glass substrates following the same procedure in device fabrication described above and Cu electrodes with 80 nm thickness were then thermal evaporated onto the resultant C 60 -SAM films through a shadow mask. C 60 -SAM films before and after crosslinking were scratched off the substrates for FTIR (Fourier transform infrared spectroscopy) analysis. The FTIR spectra of C 60 -SAM powder were collected in the transmittance mode on the PerkinElmer IR spectrometer instrument in the 400-4,000 cm À 1 region. The SEM images were taken from a Quanta 200 FEG environmental scanning electron microscope. High resolution transmission electron microscopy analysis was carried out on a FEI OSIRIS microscope. The cross-sectional samples were prepared by using focused-ion-beam equipment (FEI Helios 660).
Data availability. The data that support the findings of this study are available from the corresponding author upon request.