Controlling Crystallization of Quasi‐2D Perovskite Solar Cells: Incorporating Bulky Conjugated Ligands

Quasi‐2D hybrid halide perovskites have drawn considerable attention due to their improved stability and facile tunability compared to 3D perovskites. The expansiveness of possibilities has thus far been limited by the difficulty in incorporating large ligands into thin‐film devices. Here, a bulky bi‐thiophene 2T ligand is focused on to develop a solvent system around creating strongly vertically‐aligned (2T)2(MA)6Pb7I22 (n = 7) quasi‐2D perovskite films. By starting with a poorly coordinating solvent (gamma‐butyrolactone) and adding a small amount of dimethylsulfoxide and methanol, it is found that vertical orientation and z‐uniformity is greatly improved. These are carefully examined and verified using grazing‐incidence wide‐angle X‐ray scattering analysis and advanced optical characterizations. These films are incorporated into champion solar cells that achieve a power conversion efficiency of 13.3%, with a short‐circuit current density of 18.9 mA cm‐2, an open‐circuit voltage of 0.96 V, and a fill factor of 73.8%. Furthermore, the quasi‐2D absorbing layers show excellent stability in moisture, remaining unchanged after hundreds of hours. In addition, 2T is compared with the more common ligands butylammonium and phenylethylammonium in this solvent system to develop heuristics and deeper understanding of how to incorporate large ligands into stable photovoltaic devices.


Introduction
Hybrid organic-inorganic halide perovskite solar cells have seen tremendous growth and interest over the past 10 years, making strong strides in boosting power conversion efficiency (PCE) to rival singlejunction silicon cells. [1][2][3] The detrimental issue with 3D halide perovskite layers is that they suffer severe degradation with exposure to moisture, rendering the photovoltaic performance abysmal. [4,5] To overcome this, several efforts have been devoted to incorporating bulky and hydrophobic organic cations into the perovskite structure. Particularly, Ruddlesden-Popper 2D Perovskites (RPPs) with chemical formula L 2 A n-1 Pb n I 3n+1 , where L is an amine-terminated ligand that is larger than the Goldschmidt Tolerance and n represents the number of octahedral Pb-I layers, have been heavily investigated. [6][7][8][9][10][11][12] While improving moisture stability, the decreased electrical conductivity of the organic layers adds complexity to the perovskite absorbing layer as now crystallite/film orientation becomes critical to creating a working device. [13] Several efforts have been made to better understand and control the vertical orientation of quasi-2D films by using hot-casting, [14][15][16][17][18][19][20][21][22][23] additives, [24][25][26][27][28][29][30][31] solvent engineering, [15,[21][22][23] post annealing, [32,33] or a combination of these methods. While efficiencies for these methods have resulted in high-quality photovoltaics, the majority of quasi-2D solar cells are made with butylammonium (BA), phenylethylammonium (PEA), and their derivatives, [34][35][36][37] thus far having very few explorations into larger ligands. This is an issue as although BA and PEA RPPs have increased moisture stability over 3D perovskites, they are still susceptible to moisture degradation. Previously, our lab found that incorporating a bulky and rigid bithiophene unit (2T) into the perovskite structure gives films superior moisture stability to BA, [38] making 2T-based quasi-2D perovskite an ideal candidate for stable and efficient photovoltaics. However, there has been very limited investigations incorporating such bulky and rigid ligands into quasi-2D perovskite solar cells because controlling the nucleation and growth of such thin film is challenging.
It has been shown that to improve the vertical orientation of quasi-2D perovskite films suppression of solvate phases in the bulk solution is crucial. [39,40] Furthermore, the use Quasi-2D hybrid halide perovskites have drawn considerable attention due to their improved stability and facile tunability compared to 3D perovskites. The expansiveness of possibilities has thus far been limited by the difficulty in incorporating large ligands into thin-film devices. Here, a bulky bi-thiophene 2T ligand is focused on to develop a solvent system around creating strongly vertically-aligned (2T) 2 (MA) 6 Pb 7 I 22 (n = 7) quasi-2D perovskite films. By starting with a poorly coordinating solvent (gamma-butyrolactone) and adding a small amount of dimethylsulfoxide and methanol, it is found that vertical orientation and z-uniformity is greatly improved. These are carefully examined and verified using grazing-incidence wide-angle X-ray scattering analysis and advanced optical characterizations. These films are incorporated into champion solar cells that achieve a power conversion efficiency of 13.3%, with a short-circuit current density of 18.9 mA cm -2 , an open-circuit voltage of 0.96 V, and a fill factor of 73.8%. Furthermore, the quasi-2D absorbing layers show excellent stability in moisture, remaining unchanged after hundreds of hours. In addition, 2T is compared with the more common ligands butylammonium and phenylethylammonium in this solvent system to develop heuristics and deeper understanding of how to incorporate large ligands into stable photovoltaic devices.
of poorly coordinating solvents such as gamma-butyrolactone (GBL) and slow solvent-removal rates can increase the degree of orientation. [41,42] By using these concepts as a foundation, herein, we report a solvent engineering method utilizing a high-vapor pressure co-solvent (methanol) to induce rapid vapor-liquid nucleation while a GBL/dimethylsulfoxide (DMSO) base solvent suppresses nucleation and growth in the bulk solution, allowing for slow perovskite film growth during spin coating. With this method, we were able to create (2T) 2 (MA) 6 Pb 7 I 22 quasi-2D films with a high degree of vertical orientation, leading to devices with a PCE of 13.3%, a record for large-ligand quasi-2D perovskites. Through this method, we have created a method to create stable, vertically oriented 2T-based quasi-2D perovskite films suitable for solar cells. Furthermore, we performed a comprehensive comparison of BA, PEA, and 2T in this solvent system to better understand the relationship between the system constituents and ligand-effect on film orientation, allowing for valuable heuristics for large-ligand incorporation into quasi-2D perovskite photovoltaics.

Solvent System
The thiophene-based conjugated ammonium ligand 2T was used as a demonstrative candidate for the quasi-2D perovskite layers after it has been shown to have strong moisture-protecting qualities and significantly reduce anion diffusion when incorporated into the perovskite structure. [38] (2T) 2 (MA) 6 Pb 7 I 22 (Figure 1A) was used as the main study material due to decreasing band gap and increased device performance. The chemical structures for 2T, as well as the more commonly used PEA and BA are shown in Figure 1B. Gamma-butyrolactone (GBL), a poorly-coordinating solvent, was used as the main solvent as it has been used for single crystal growth of perovskite materials and has been demonstrated to be beneficial for the vertical alignment of the quasi-2D domains. [41] The low supersaturation conditions with GBL, allows for crystal growth to occur more rapidly than nucleation, leading to larger crystallite sizes. The  6 Pb 7 I 22 . B) Chemical structure of different ligands that can be used to make quasi-2D perovskite films. C) Schematic for the proposed mechanism of thin film growth with the hydrophobic and bulky 2T ligand. Inorganic Pb/I/MA segments are shown in blue and 2T organic ligand is shown in green. Fast surface nucleation leads to better phase-purity and z-uniformity and low bulk nucleation leads to highly oriented films. Both are required for suitable photovoltaic absorbing layers. effect GBL has on aiding vertical orientation with slow solvent evaporation has been demonstrated with BA and PEA ligands, although with diminishing effect as ligand length and degree of conjugation increases. [41] As discussed later, this is most likely due to often overlooked kinetic differences between the cations. The diminishing effect on larger cation orientation leads to GBL alone only being able to partially orient the 2T ligand perovskites and leading to films with poor z-uniformity. To remedy this, we propose a general mechanism that includes the common idea to suppress bulk nucleation and also add the stipulation that for larger ligands, rapid surface nucleation should occur. Suppressing bulk nucleation aids in vertical orientation whereas rapid surface nucleation helps improve phase-distribution and z-uniformity, as shown in Figure 1C. To suppress bulk nucleation, we found by adding a small amount of dimethylsulfoxide (DMSO) into GBL (10 vol%) we could greatly increase the vertical orientation. To help induce rapid surface nucleation, we add methanol (MeOH) as a co-solvent, which has a much higher vapor pressure than GBL, and evaporating more quickly during the spin coating process. This fast evaporation, we theorize, induces a local-supersaturation at the vapor-liquid interface and causes rapid-nucleation. Having rapid surface nucleation without suppressed bulk nucleation will lead to films with improved z-uniformity, however with poor device performance due to more random crystal domain orientation. Only with the combination of both fast surface nucleation and suppressed bulk nucleation can we generate ideal quasi-2D absorbing layers for efficient photovoltaic devices. Focusing on these two concepts independently can give us deep insights to how to engineer a solvent system to create high-efficiency perovskite layers with larger ligands.

Thin Film Morphology and Structure
In order to study how the solvent engineering effect manifests in solar cells, we first turned to study the effects on the perovskite film quality, morphology, and structure by using (2T) 2 (MA) 6 Pb 7 I 22 films formed by spin coating on glass substrates. To probe the thin film crystallite orientation, grazingincidence wide-angle X-ray scattering (GIWAXS) analysis was used. An incident angle (α) was chosen as 0.3° to be able to get sufficient diffraction intensity from the bulk of the film. The diffraction pattern remains consistent through different α values ( Figure S1, Supporting Information) with the exception of α = 0.1°. This is explained by understanding the mechanism of crystallization of quasi-2D films at the liquid-vapor surface, which prefer vertical orientation. [13] However, if bulk nucleation is not sufficiently suppressed, this can lead to anisotropies at different X-ray probing depths. For the control case of pure GBL for one-step spin coating, the vertical orientation is poor as evidenced by GIWAXS spectra in Figure 2A. The Debye-Scherrer rings at q ≈ 1 Å -1 and 2.0 Å correspond to the (111) and (202) planes of the perovskite crystal respectively, films and can be used as an overall indication of film orientation. [14] By the addition of a small amount of DMSO (10 vol%), the Debye-Scherrer rings give way to Bragg Spots, suggesting improved crystallinity. The lack of diffraction rings while pronounced diffraction spots along the 0° azimuthal angle in the spectra gives an indication of increased out-of-plane orientation. Methanol addition into the precursor solution seems to have little effect on film orientation, with GBL/MeOH and GBL/DMSO/MeOH showing diffraction patterns that are similar to their GBL and GBL/ DMSO solutions, respectively. When MeOH is employed, we can also see a decrease in diffraction intensity corresponding to low n-number phases between q = 0.25 and 0.5 Å -1 , indicating a higher amount of preferable high-n phases. Furthermore, when MeOH is utilized in conjunction with GBL and DMSO, there is an increase in intensity for the diffraction spots, indicating an increased crystallinity with MeOH addition. We can quantify the degree of orientation by analyzing the intensity along a fixed q-value as a function of the azimuthal angle. Taking the (111) plane (q xy = 1.0 A -1 ) as an example and plotting the logarithmic intensity along different azimuth angles, we can see the effect each solvent system is having on orientation, as seen in Figure 2B. The systems with DMSO greatly increase vertical orientation, lowering intensity between 20° and 80° by orders of magnitude. As a note, without DMSO, there is still partial vertical orientation observed, with the strongest intensity at an azimuthal angle of 0°, however, a much stronger vertical orientation is required for devices.
Furthermore, when looking at the X-ray diffraction (XRD) pattern of the perovskite films ( Figure 2C) the addition of DMSO decreased the full-width half max (FWHM) of the (111) peak. The addition of MeOH decreased the FWHM even further, with the addition of both DMSO and MeOH leading to a minimization of the diffraction peak. The average crystallite size can be calculated by using Scherrer's Equation D = Kλ/ (β cosθ) where D is the average crystallite domain size, K is the shape factor (taken here to be 0.89), λ is the X-ray wavelength (0.15 406 nm) β is the FWHM taken form XRD spectra, and θ is the Bragg angle. The effect the different solvent conditions have on FWHM is shown in Figure 2D. The solvent system of GBL/DMSO/MeOH increases the average crystallite domain size from 35 nm (GBL only) to 52 nm (GBL/DMSO/MeOH), resulting in a more crystalline film. In addition, the GBL/ DMSO/MeOH films are ideal for solar cells due to far fewer pin-holes ( Figure S2, Supporting Information), which is important for reducing shunting.
Contrasting these results with using the same stoichiometric ratios in the polar solvent dimethylformamide (DMF), the 2T perovskite films lead to strongly disordered films ( Figure S3, Supporting Information). When spin-coated in pure DMF, the films show a GIWAXS spectra that feature strong Debye-Scherrer rings, indicating poor orientation. In addition, there are several other diffraction peaks that show up that we attribute to lower n-number phases that are formed in the film. Even with the addition of MeOH, the diffraction pattern for films in DMF has no difference. DMF has been employed heavily to lead to vertically oriented BA films. [14,41,[43][44][45] This gives a stronger understanding that for the larger conjugated ligands, other base solvents should be employed in order to achieve preferred orientation.
While diffraction techniques can give an indication of film crystallinity and orientation, phase distribution cannot be determined as easily. In order to better understand the composition of the formed thin film, optoelectronic characterizations were used. The photoluminescence (PL) from the topside of the film versus the bottom can give a good indication of z-direction uniformity, or phase uniformity as a function of film depth. [32] The schematic for this characterization is seen in Figure 3A. Looking at the normalized PL spectra for each solvent condition in Figure 3C there is a very prominent peak ≈750 nm for each film, which corresponds to high n-phase emission. [46] Comparing the normalized top and reverse excitation for the GBL and GBL/DMSO conditions, there is a clear anisotropy in emissions. The reverse excitation has strong peaks at 568, 610, and 660 nm which correspond to emissions from n < 5. The top excitation has emissions at the same energy levels, but at a reduced magnitude, indicating there is a phase mismatch throughout the film with more low n-phases being formed on the bottom of the film (i.e., the film/substrate surface). This observation is consistent with the hypothesis of how quasi-2D film formation occurs: high n-number phases form at the liquid/surface interface and propagate downwards. [39] For the one-step spin-coating done here, all precursors (MAI, PbI 2 , 2TI) are mixed together based on desired n-number (i.e., n = 7). If larger n-values are formed at the liquid-vapor surface, this means that the ratio of MAI/2TI will decrease as crystallization continues. This will lead to lower n-domains forming on the liquid-substrate surface due to precursor availability and excess 2TI. Without DMSO, the films still seem to form low n-phases, as seen by the absorption spectra in Figure 3B however with the addition of DMSO and MeOH, suppression of low n-number phases is seen.
Comparing the solutions with MeOH introduced, we see that the front and back excitations are similar, indicating the MeOH improves the z-uniformity of the quasi-2D film. Furthermore, the lower n-number emission peaks are suppressed with the MeOH addition, indicating more consistent phase-distribution. The decreased solubility of 2TI in MeOH ( Figure S4, Supporting Information) indicates that the suppression of lower n-phases is most likely not due to preferential solubilization of MAI versus 2TI, as we would expect to see much higher levels of low n-phases with the addition of MeOH, whereas the opposite is observed.
To better isolate the emissions in the quasi-2D films confocal photoluminescence was used. As seen in Figure 3D, a range of 540-670 nm was picked to isolate the lower n-number emissions (n < 5). In pure GBL, the perovskite film has a weak emission from 544-571 nm, whereas all other solvent systems did not. This can indicate in the poor-coordinating solvent GBL, significant lower n-number domains are formed. This can be explained by the reactivity difference between 2TI and MAI. As GBL slowly evaporates away during the spin coating process, MAI will coordinate with the PbI 6 octahedra preferentially compared to 2TI, leading to higher n-domains at the surface, causing lower n-domains to form as crystallization continues. Emissions between 571-597 nm are observed in both the GBL and GBL/DMSO solvent systems, whereas these low n-number emissions are suppressed in the GBL/MeOH and GBL/DMSO/MeOH systems.
Taking the diffraction and optoelectronic characterization results together we can better understand the role of each solvent added to the system. GBL as a poor-coordinating solvent leads to partially-oriented films with poor z-uniformity and significant amounts of low n-domains. When DMSO is added, vertical orientation improves greatly, however, phase uniformity throughout the film remains poor, and low n-number emissions can still be observed. It was found that adding MeOH leads to partially-oriented films similar to the pure GBL case, however, differs in that there is improved z-uniformity with suppression of low n-number emissions. Interestingly, the combination of GBL/DMSO/MeOH incorporates the benefits of both DMSO and MeOH leading to improved vertically-aligned films with improved z-uniformity and suppression of low n-number emissions. This result is also echoed when time-of-flight secondary ion mass-spectrometry (ToF-SIMS) was performed on the films in Figure S5, Supporting Information. Tracking the ratio of I/Pb (Cs 2 I + /CsPb + ) allows us to have an indication of n-number throughout the film. As seen in Figure S5, Supporting Information, the solutions with MeOH in them have a more consistent I/Pb ratio as compared to solutions without, indicating that there is a more consistent n-number throughout the z-direction of the film.

Photovoltaic Device Performance
After characterizing the 2T n = 7 films and identifying the solvent system that allows for improved phase purity, z-uniformity, and vertical orientation, photovoltaic devices were fabricated to understand how these absorbing layers perform in an actual device configuration. The devices were fabricated using the inverted architecture of ITO/PEDOT:PSS/PVSK/PCBM/BCP/ Ag as seen in Figure 4A. After optimizing the devices based on anneal time and temperature ( Figures S6 and S7, Supporting Information), ideal anneal conditions were found to be 65 °C for 60 min. The current-voltage curves each solvent system are shown in Figure 4B. The champion pure GBL case led to devices with a PCE under 4%. With the addition of only MeOH, device performance actually decreases, with barely functional devices. The GBL/DMSO system was optimized to give devices with a max efficiency of 6.33%, with a short-circuit current density (J SC ) of 9.78 mA cm -2 , an open circuit voltage (V OC ) of 0.90 V, and a fill factor (FF) of 71.4%. Comparing both the GBL and GBL/MeOH systems to the GBL/DMSO it is clear that vertical orientation is the most influential parameter for increasing device performance. The combination of GBL/DMSO/MeOH was optimized to give devices with a max efficiency of 13.3%, with a doubled J SC of 18.9 mA cm -2 , a V OC of 0.96 V, and a fill factor of 73.8%. As a note, this is near double of previously reported literature using a similar ligand. [32] Furthermore, the hysteresis of the perovskite solar cells is greatly improved with the addition of DMSO and MeOH, giving a hysteresis index [42] of 0.01, compared to 0.06, 0.40, and 0.32 with GBL/DMSO, GBL/MeOH, and GBL respectively. This trend is maintained for many batches of devices as well as for champion devices, as seen in Figure 4C and efficiencies for the GBL/DMSO/MeOH case are stable under consistent bias ( Figure S8, Supporting Information). Figure 4D shows the external quantum efficiency (EQE) of the 2T-based devices with different solvent conditions. It is clear that the addition of DMSO and MeOH into the GBL system increase the absorption across all wavelengths, with reaching near 80% over the wavelength range from 400-600 nm leading to the increase in current and improving the device efficiency. The integrated J SC gives 15.451 mA cm -2 , which is within 0.5 mA cm -2 of the measured 16 mA cm -2 of the device. The EQE of the GBL only case shows low efficiency overall, most likely due to the in ability for generated charge carriers to be extracted out of the absorbing layer because of partial vertical orientation. For the GBL and DMSO solvent system, the absorption is decreased for all wavelengths and has an integrated J SC of 9.717 mA cm -2 which is within 0.2 mA cm -2 of the measured 9.9 mA cm -2 . The peak at around 400 nm we attribute to low n-phases that form from only using GBL, as discussed earlier in this article. The GBL/MeOH devices have little to no charge extraction and so have abysmal EQE efficiency. Figure 4E  double-logarithmic plot based on the expression J SC ∝ I α , where I is the incident light. A value of α closer to unity indicated the photovoltaic device has little space charge limited behavior. [24] For the GBL/DMSO/MeOH case and GBL case, α is over 0.9, while GBL/DMSO is only 0.765. This indicates that although GBL/DMSO improves vertical orientation, there is a trade-off with increased bimolecular recombination. [27] Since the incorporation of bulky organic ligands is driven by stability, we monitor the long-term performance of the quasi-2D films. The fabricated 2T-based devices have excellent intrinsic stability when left in an inert environment, having identical current-voltage performance after 60 days ( Figure S9, Supporting Information). Films were used for humidity testing, as PEDOT:PSS is hygroscopic and will lead to degradation of the device on a similar timeframe regardless of the ligand used ( Figure S10, Supporting Information). PTAA was investigated to be used to circumvent this, however, it was observed that the crystallization is highly surface-sensitive, leading to poor morphology when using PTAA ( Figure S11, Supporting Information). Performing stability tests on spin coated films, 2T quasi-2D films show no degradation after 500 h in a 65% relative humidity (RH) environment ( Figure 4E), far outpacing BA and PEA films ( Figure S12, Supporting Information) The incorporation of the bulky conjugated ligands of 2T give the perovskite strong stability and high moisture resistance.
Furthermore, alcohols with lower vapor pressure were investigated, and it was found that vapor pressure has a positive relationship with device efficiency, with lower vapor pressure co-solvents such as benzyl alcohol and n-butanol having worse device characteristics than ethanol or methanol ( Figure S13, Supporting Information). An explanation for this is that MeOH helps induce surface nucleation from its rapid evaporation compared to the base solvent GBL, however when the boiling point of the co-solvent is more similar to GBL, the effect of having a co-solvent to help nucleation diminishes.

Vertical Orientation Trends
To vertically align perovskite films, the idea has been to suppress bulk-film nucleation so that nucleation only occurs at the liquid/vapor surface and propagates downwards. [13,39] The main conclusion from previous work is that there should be a stark difference between the rates of nucleation on the surface of the film versus the bulk of the film. Performing a simple test with n = 1 perovskite films (no MA + ), we have found that the rates of formation are dependent on the ligand being used ( Figure S14, Supporting Information). This was found by spin coating films of L 2 PbI 4 (L = . BA, PEA, or 2T) and using an in-situ photoluminescence heating stage to characterize the formatting of the perovskites. This rapidity of formation for smaller ligands such as butylammonium (BA) may be partially to explain why they have had such a dominance of high-efficiency quasi-2D PVSCs in the past 5 years ( Figure S15 and Table S1, Supporting Information). With our solvent system of GBL/DMSO/MeOH, the DMSO helps the suppression of bulk nucleation leading to vertical alignment, while the high vapor-pressure of the MeOH helps induce nucleation at the surface during the spin coating procedure aiding in phase purity and z-uniformity. While this solvent system works well for the bulky 2T ligand, we wanted to investigate how two commonly used ligands, butylammonium (BA) and phenylethylammonium (PEA) would compare. By using a static GIWAXS technique, the diffraction spectra are shown in Figure 5, along with the azimuthal-dependent intensity of the (111) plane. Looking at the BA films, high levels of orientation are observed in all conditions, indicating the BA system is more robust when it comes to film formation. When moving to conjugated ligands, the solvent system has much more of a role. The PEA and 2T ligands have similar low out-ofplane orientation in the GBL and GBL/MeOH systems, where we can see very strong diffraction rings for the (111) plane. With the addition of DMSO into the system, we see similar behavior with PEA as with 2T, with improved vertical orientation as shown by sharp diffraction spots. Although smaller in magnitude, the GBL/DMSO/MeOH condition across all ligands appears to enhance vertical orientation as well as improve crystallinity as demonstrated by increased intensity when compared to the GBL/DMSO system alone, giving insight into the role of MeOH even further.

Conclusion
To summarize, we have developed a method to create vertically oriented, z-uniform, quasi-2D perovskite thin films with large bulky ligands, using 2T as a focus. To achieve this, we created a solvent system that takes the existing idea of suppressing bulk nucleation, and add the requirement to enhance vaporliquid surface nucleation by addition of a high vapor pressure co-solvent. The suppressed bulk nucleation is aided by a small amount of DMSO leads to highly oriented films, whereas the rapid surface nucleation induced by MeOH leads to improved z-uniformity and phase purity. We demonstrated this approach by creating quasi-2D 2T-based absorbing layers with high stability in moisture environments, leading to the fabrication of solar cells based on the 2T ligand with a PCE of over 13%. Furthermore, we perform a comprehensive GIWAXS study comparing 2T to the more common BA and PEA ligands to give insight to requirements for creating large-ligand quasi-2D films suitable for photovoltaics. With this novel approach to solvent engineering, we generate a framework for large-ligand quasi-2D perovskite films that stresses the importance of controlling nucleation and crystal growth, which has been difficult for large bulky ligands up until now.
Fabrication of Perovskite Photovoltaics: Pre-cut glass/ITO substrates were washed with soap and water, deionized water, acetone, isopropanol, acetone, and then isopropanol in an ultrasonic bath for 20 min for each step. The substrates were treated with UV-ozone for 15 min. PEDOT:PSS (Ossila) is extruded through a 0.2 µm filter, spin-coated in ambient conditions at 5000 rpm, and then annealed at 120 °C for 20 min. The substrates are quickly taken into a nitrogen-filled glovebox where the perovskite later will be spin-coated. The perovskite layers were spincoated at 4000 rpm for two minutes, then annealed at 65 °C for 90 min. After the substrates return to room temperature, a 20 mg mL -1 solution of PCBM in chlorobenzene is spin-coated at 2000 rpm for 30 s and then annealed at 70 °C for 2 min. A 0.5 mg mL -1 solution of BCP in isopropanol is then spin-coated at 2000 rpm for 30 s and then annealed at 70 °C for 2 min. Finally, 90 nm of silver is thermally evaporated onto the devices at <4 e-6 torr.
Characterization: J-V characteristics were obtained using a Keithley 2450 source meter with a simulated AM 1.5G irradiation (100 mW cm -2 ) from a xenon-arc lamp (Enlitech SS-F5-3A) that was calibrated with a Si solar cell (Enlitech). The device area was 0.044 cm 2 , defined by shadow mask during thermal evaporation. The devices were measures from 1.2 V to −0.2 V then back to 1.2 V. Powder diffraction (PXRD) was performed to confirm the crystal structure of samples. The PXRD was done by a Rigaku SmartLab Diffractometer. Single-crystal XRD was performed to resolve the atomic structure of the materials and was done using a Bruker AXS D8 Quest CMOS diffractometer. GIWAXS spectra will be taken at beamline 7.3.3. at the advanced light source (ALS) at Lawrence Berkeley National Lab utilizing an incident wavelength of 10 keV. 2D spectra can be recorded with a Pilatus 2M-2D detector and integrated to reduce to 1D with the NIKA GIWAXS software. Positive high mass resolution depth profiles were performed using a TOF-SIMS NCS instrument, which combines a TOF-SIMS instrument (ION-TOF GmbH, Münster, Germany) and an in-situ Scanning Probe Microscope (NanoScan, Switzerland) at Shared Equipment Authority from Rice University. The analysis field of view was 100 × 100 µm 2 (Bi 3 + @ 30 keV, 0.3 pA) with a raster of 128 by 128 along the depth profile. A charge compensation with an electron flood gun has been applied during the analysis. An adjustment of the charge effects has been operated using a surface potential. The cycle times was fixed to 100 µs (corresponding to m/z = 0 -911 a.m.u mass range). The sputtering raster was 450 × 450 µm 2 (Cs + @ 1 keV, 45 nA). The beams were operated in non-interlaced mode, alternating 1 analysis cycle and 1 frame of sputtering (corresponding to 1.55 s) followed by a pause of 2 s for the charge compensation. The MCs n + (n = 1, 2) depth profiling has been also used for improving the understanding of the data. [47][48][49][50][51] This is a useful method, mainly applied to quantify the alloys but also to identify any ion compounds. The cesium primary beam is used for sputtering during the depth profile and permits to detect MCs + or MCs 2 + cluster ions where M is the element of interest combined with one or two Cs atoms. The advantages of following MCs + and MCs 2 + ions during ToF-SIMS analysis include the reduction of matrix effects and the possibility of detecting the compounds from both electronegative and electropositive elements and compounds. All depth profiles have been point-to-point normalized by the total ion intensity and the data have been plotted using a 10-points adjacent averaging. Both normalization and smoothing have permitted a better comparison of the data from the different samples. The depth calibrations have been established using the interface tool in SurfaceLab version 7.2 software from ION-TOF GmbH to identify the different interfaces and based on the measured thicknesses using DEKTAK profiler. The SEM images were collected using a Hitachi 4600-S scanning electron microscope. A 10 kV electron beam was used to collect secondary electron signals. The transmittance of thin films was measured by a Cary 60 Ultraviolet-visible light spectrophotometer, used in transmission mode. The thin films on quartz were used to collect absorption spectrum data in transmission mode. Bright-field optical images were taken using a custom Olympus BX53 microscope. Photoluminescence images were collected using a fluorescent light source (012-63000; X-CITE 120 REPL LAMP). The filter cube contains a bandpass filter (330-385 nm) for excitation, and a dichroic mirror (cutoff wavelength: 400 nm) for light splitting and a filter (Long pass 420 nm) for emission.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.