Highly Efficient and Stable Wide‐Bandgap Perovskite Solar Cells via Strain Management

Wide‐bandgap (WBG) perovskite solar cells (PSCs) with high performance and stability are in considerable demand to boost tandem solar cell efficiencies. Perovskite bandgap broadening results in a high barrier for enhancing the efficiency of PSCs and phase segregation in perovskite. In this study, it is shown that the residual strain is the key factor affecting the WBG perovskite device efficiency and stability. The dimethyl sulfoxide addition helps lead halide with opening the layer spacing to form intermediate phases that provide more nucleation sites to eliminate lattice mismatch with organic components, which dominates the strain effects on the WBG perovskite growth in a sequential deposition. By minimizing the strain, 1.67 and 1.77 eV nip devices with record efficiencies of 22.28% and 20.45%, respectively, can be achieved. The greatly suppressed phase segregation enables the devices with retained 90–95% of initial efficiency over 4000 h of damp stability and 80–90% of initial efficiency over 700 h of maximum‐power‐point (MPP) stability. Besides, the 1.67 eV pin devices can achieve a competitive 22.3% efficiency with considerable damp‐heat, pre‐ultraviolet (pre‐UV) aging and MPP tracking stability according to IEC 61215. The final efficiency of more than 28.3% for the perovskite/Si tandem is obtained.


Introduction
As the most promising next-generation photovoltaic technology, perovskite solar cells (PSCs) have achieved a record efficiency of 25.7%; these PSCs mainly comprise Formamidinium lead iodide (FAPbI 3 ) with a bandgap of ≈1.5 eV. [1][2][3] Most PSCs with efficiencies over 23% employ perovskites with a bandgap below 1.6 eV containing FAPbI 3 as main compositions and a fraction of compositions containing bromide (Br) and cesium (Cs), or even none. [2][3][4][5][6] Br addition can broaden the bandgap of perovskites while optimal Cs introduction is evidenced to boost the device performance and stability. [7,8] With the efficiency approaching to the Schottky limit, a low-bandgap material such as Si, Pb-Sn perovskite and CIGS can be used to construct a tandem for further increasing the efficiency. [9][10][11] Efficient tandems require efficient PSCs with an optimal bandgap generally greater than 1.65 eV. [12,13] However, with the increased Br and Cs contents in perovskite, the bandgap increases while the efficiency and stability are limited mainly due to severe nonradiative recombination and phase segregation in perovskites. [14,15] The highest efficiency of the 1.68 eV WBG perovskite, which is widely applied in perovskite/Si 2-T tandems has exceeded 20.5%, resulting in a final efficiency of over 29% in a tandem. [9,12,16] For all perovskite tandem systems, the optimal 1.77 eV bandgap PSCs delivered the highest efficiency of ≈16.5-17.5% due to the severe nonradiative recombination. [10,17,18] The performance enhancement of WBG PSCs is much slower than that of sub-1.6 eV PSCs; this hinders the development of perovskite-based tandems.
For the WBG perovskite, a thermodynamically stable Brrich phase preferentially forms in the precursor solutions and films, resulting in multiple nonradiative recombination both in bulk perovskite and interfaces with transport layers. [19,20] Halide migration can easily occur in defective perovskites with halide vacancies and grain boundary passages, resulting in phase segregation. [21,22] Hence, defect engineering is one of the most commonly adopted strategies to mitigate phase segregation. [23,24] The partial substitution of the A site group by K + , B site group by Sn 2+, and X site group by Cl − or SCN − are known to improve the grain growth and eliminate defects to realize high-quality perovskites. [25][26][27][28] Yang et al. utilized tribromide to obtain customized defects in a 1.65 eV perovskite and achieved a high open circuit voltage (V OC ) over 1.2 V; they finally prepared a perovskite/Si tandem with 28.6% efficiency and good stability. [29] Surface passivation can indeed improve the performance of the WBG perovskite devices, a few of which can greatly enhance the device stability mostly based on the phase stability. [30,31] Each additional interface implies one more parasitic absorption, which is unfavorable for developing tandem solar cells.
Few efforts were made toward intrinsic carrier recombination management in WBG perovskite, demonstrating that such studies remain a challenge. Regardless, the strain in the perovskite is known to be a key factor that affects the induced defects, ion migration, component segregation, and lattice stabilization in perovskite, further dictating device's efficiency and stability. [32][33][34] Yang et al. confirmed that micro strain in MAPb(I 0.6 Br 0.4 ) 3 causes compositional inhomogeneity and phase segregation. By partially replacing Pb 2+ with Sn 2+ , the ion migration barrier can be increased and thus phase segregation can be mitigated. [28] Wang et al. added adenosine triphosphate into 1.65 eV perovskite to modulate tensile strain and realized an illumination-stable nip perovskite/Si tandem solar cell. [35] However, the excellent tandem solar cells rely on efficient and stable perovskites with larger bandgaps (≈1.7 eV) which suffer from severer performance loss and phase segregation.
Here, we developed a sequential method to fabricate WBG perovskite solar cells with state-of-the-art efficiencies of ≈22.5% for 1.67 eV nip devices and ≈20.3% for 1.77 eV nip devices, and ≈22.3% for 1.67 eV pin devices. Strain analysis based on X-ray diffraction (XRD) patterns and Raman spectra were conducted for the film depositions. Careful control of layered-lead-halideopening induced strain by DMSO in the precursor solution enables the transition from tensile strain to compressive strain in the perovskites. The strain-free WBG perovskite with exhibits the best phase stabilization under light soaking. The resultant nip devices have over 4000 h of storage stability and over 700 h of maximum-power-point (MPP) output stability under fullspectrum light without encapsulation. The resultant pin devices presented excellent performances in damp-heat, pre-ultraviolet (pre-UV) aging and MPP tracking stability tests according to IEC 61215. By integrating these pin configurations on Si subcells, an efficiency of over 28% can be achieved, demonstrating the valuable applications of the strain-free WBG PSCs.

Strain Characterization and Analysis
In perovskite fabrication, methods are applied to first dissolve PbX 2 (X = I, Br, Cl) using DMF/DMSO and then to dissolve MAX or FAX using isopropanol (IPA). For WBG perovskites, it is necessary to add cesium compounds to adjust a bandgap and obtain high-quality perovskites with reduced trap densities and charge recombination rates. [36] Thus, CsX is introduced in the first deposition to form (CsI) 0.08 (PbI 1.4 Br 0.6 ) for 1.67 eV perovskite or (CsI 0.125 Br 0.875 ) 0.08 (PbI 1.2 Br 0.8 ) for 1.77 eV perovskite films owing to its insolubility in IPA. The surface scanning electronic microscopy (SEM) morphologies of (CsI) 0.08 (PbI 1.4 Br 0.6 ) and (CsI 0.125 Br 0.875 ) 0.08 (PbI 1.2 Br 0.8 ) films with different DMSO contents are shown in Figures S1 and S2 (Supporting Information). The DMSO addition drastically enhances dendrite growth because of the supersaturation of the mesophase, which is shown in Figure S3 (Supporting Information) based on conventional XRD imaging results. [19] Compared to the XRD results of bare powders without DMSO treatment in Figure S4a (Supporting Information), peaks were newly generated at 8°-11°, which correspond to DMSO-related intermediate phases. [37,38] We carefully examined the black-circled PbI 2 peak position in Figure S4a (Supporting Information); the details are shown in Figure S4b (Supporting Information). The peak position shifts to smaller angles due to the DMSO addition, which indicates that the d spacing of PbI 2 increases probably with the strain effect. Furthermore, we conducted Fourier transform infrared spectroscopy measurements to verify the existence of DMSO, as shown in Figure S5 (Supporting Information). The overall peak position is basically consistent with the peak position in the DMSO IR spectrum. For the circled regions, which correspond to the SO vibration and sp [3] CH bonding (DMSO fingerprint signals), DMSO clearly remains in the film. The d spacing determined through XRD with stress attachments is examined to determine the strain influence on (CsI) 0.08 (PbI 1.4 Br 0.6 ) and (CsI 0.125 Br 0.875 ) 0.08 (PbI 1.2 Br 0.8 ) films, as shown in Figure 1a,b. According to the Bragg equation, the d-spacing increases as the azimuth angle (ψ) of the crystal plane increases. The linear fit of 2θ versus sin 2 ψ illustrates the enhanced tensile strain for (CsI) 0.08 (PbI 1.4 Br 0.6 ) and (CsI 0.125 Br 0.875 ) 0.08 (PbI 1.2 Br 0.8 ) films, as shown in Figure 1c,d. [39,40] The graphs of the tensile strain development with annealing temperature, as shown in Figures S6-S11 (Supporting Information), illustrate that a higher temperature induces a larger tensile strain, which is correlated with the coefficient of thermal expansion (CTE) mismatch model. [41,42] The corresponding strain calculation formula is discussed in the Experimental Section. DMSO enters the atomic layers of lead halide and stretches the layers vertically or laterally to form a mesophase, which can be identified in the Raman spectra (Figure 1e,f). [43][44][45] The vertical arrangement is more conducive to the growth of vertically penetrating perovskite grains, which contributes to carrier transport. There are two signals that correspond to the laterally arranged lead halide at ≈95 cm −1 and vertically arranged lead halide at ≈110 cm −140 . Introduction of DMSO reduced the contents of the laterally arranged (CsI) 0.08 (PbI 1.4 Br 0.6 ) and (CsI 0.125 Br 0.875 ) 0.08 -(PbI 1.2 Br 0.8 ) intermediates, and the Raman peaks of both films shifted to a smaller wavenumber, indicating a tensile strain as proved by the XRD results. [46,47] With the further deposition of MAX/FAX dissolved in IPA, two WBG perovskite films with different bandgaps, as confirmed by the UV-vis absorption spectra in Figure 2a,b, were fabricated successfully. The absorption edges of these films are slightly different than those of films with the same perovskite composition due to the strain; the bandgaps of the two perovskite films are ≈1.67 and 1.77 eV. It is found that the transition from tensile to compressive strain can result into a red shift in the absorption edge, as will be demonstrated below. [39] The perovskite characteristics are distinguished using SEM, Atomic force microscopy (AFM), and XRD measurements. The tensilestrain induced precursor films with expanded lead hailed layers assist perovskite grain growth with DMSO content lower than 25% (Figures S12 and S13, Supporting Information). However, the crystal grains become smaller when the DMSO content is 40%, possibly because of the presence of excessive perovskite nucleation centers. The largest average grain sizes are ≈565 and www.afm-journal.de www.advancedsciencenews.com

(3 of 11)
© 2023 Wiley-VCH GmbH 650 nm for 1.67 eV perovskite and 1.77 eV perovskites, respectively, with 25% DMSO. The AFM results in Figures S14 and S15 (Supporting Information) reconfirm the growth pattern. With DMSO increase in the precursor solution, the roughness of the perovskite surface becomes larger due to the appearance of more obvious grain wrinkles, indication the transformation from tensile strain to compressive strain. [48] The XRD patterns of perovskites with different DMSO content reveals that more DMSO helps to form more intermediate phases with PbI 2 , resulting into a final transformation of perovskite with reduced residual PbI 2 (red-circled in the SEM images in Figures S12 and S13, Supporting Information). The compositional homogeneity on the perovskite surface or inside the perovskite is illustrated from the SEM-EDS line-scan images (Figures S17-S22, Supporting Information). In the absence of MAI/MABr/FAI/FABr, the signal intensity of the halogens (Br or I) will be weakened, which is in agreement with the XRD results.
An eventual strain analysis was conducted on perovskite films by considering the XRD patterns shown in Figure 2c,d. The fitting results from Figure 2e,f prove that the residual tensile strain gradually transforms to compressive strain with an increase in DMSO with hardly any tensile strain after high-temperature annealing for both perovskites. Perovskite films with 25% DMSO have almost no residual stress or only have a slight stretching tendency. Nevertheless, high annealing temperatures still make perovskites tend to stretch, as shown in Figures S23-S28 (Supporting Information); this stretching tendency promotes the transformation of the original compressive (tensile) strain to tensile (much more tensile) strain, accounting for the wrinkled perovskite surface with increased DMSO. Strain-free perovskite is less prone to generate lattice dislocation and defects (non-radiative recombination). [32,33,39] We inferred that the strain is mainly applied on FA/MA-[PbX 6 ] bonds due to the significant shift of organic cation modes from Raman results in Figure S29 (Supporting Information), showing a tendency for compression versus DMSO increasement. [49,50] In support of the strain type transition, the nanoindentation test was conducted on perovskite films with two   Figure S30 (Supporting Information). Both perovskite samples with 1.67 and 1.77 eV present the strain type from tensile to compressive with increased DMSO content, corresponding to the results from XRD and Raman measurements due to that compressive strain can increase the maximum load that the sample withstand, while tensile strain does the opposite. [51,52] The charge carrier properties of strained perovskite films on glass/ITO/SnO 2 substrates are delved using photoluminescence (PL) and time-resolved photoluminescence (TRPL) tests as shown in Figure S31 (Supporting Information). As expected, perovskites with 25% DMSO had the strongest PL intensity, which indicates the least nonradiative recombination. A significantly slower TRPL decay was recorded for strain-free perovskites with 25% DMSO in accordance with steady state PL tests. [53][54][55] Both tensile and compressive strains enable the carriers to form more nonradiative recombination centers in perovskites due to the stress effect.
Here, PbI 2 content dominants in the first deposition step of (CsI) 0.08 (PbI 1.4 Br 0.6 ) and (CsI 0.125 Br 0.875 ) 0.08 (PbI 1.2 Br 0.8 ) films with a little CsI/CsBr in DMF/DMSO (10%, 25% or 40%) on SnO 2 substrates, we choose PbI 2 as the representative for further explanation in Figure 3. Based on the bigger thermal expansion coefficient of PbI 2 than SnO 2 , and that DMSO can penetrate to open the dense PbI 2 crystal layers, tensile strain was formed at first. More DMSO penetration can stretch PbI 2 lattice larger, leading to larger tensile strain. However, with the insertion of organic cations into PbX 2 -DMSO intermediates in the second step, perovskite transition occurs along with the fast evaporation of DMSO in annealing, and compressive strain can be formed easily. [40] Higher DMSO content forms more PbX 2 -DMSO intermediates and thereby converts to perovskite more thoroughly. The final strain type is the compensation result of tensile strain formed by PbX 2 -DMSO and compressive strain formed by perovskite transition. In order to prove the mechanism, we supplemented related strain analysis using N-Methylpyrrolidone (NMP) instead of DMSO, the results are shown in Figure S31 Figure S32b,c (Supporting Information) corroborate the trend of reduced tensile stress with increased NMP content in perovskite films. All photovoltaic performance parameters generally get improved, especially for V OC and FF ( Figures S35 and S36, Supporting Information). The hysteresis also improved a lot with strain engineering. The J SC and bandgap of perovskite were calibrated by external quantum efficiency (EQE) in Figure S37 (Supporting Information), resulting in an integrated J SC of 21.25 and 19.45 mA cm −2 . The detailed PV parameters of the state-of-the-art WBG perovskite solar cells are summarized in Table S1 (Supporting Information), evidencing the breakthrough performance of our devices. The enhanced J SC and FF in this study suggest that high-quality WBG perovskite with less defects can be obtained using the subsequent method as well as strain engineering.

Performance and Stability of Nip Devices
The stability results are concluded in Figure 4g,h. The unencapsulated devices for the target 1.67 eV WBG perovskite had almost no efficiency loss for the first 700 h in a N 2 glove box, and over 95% after the next 3500 h in the damp exposure with a relative humidity of 20-60%. Similarly, the target 1.77 eV WBG PSCs with more Br and Cs retained over 90% of their initial efficiency. However, the efficiency of 1.77 eV PSCs fluctuated more drastically, possibly because of the easier phase separation and recovery. To verify the tolerance under standard 1-sun illumination, the devices were placed and traced by a solar simulator (Enli Tech.) without a filter in a nitrogen glovebox in Figure 4h and Figure S38 (Supporting Information). The target 1.67 and 1.77 eV WBG PSCs retained over 90% and 80% of their initial efficiency after 700 h MPP tracking, respectively, competitively among the existing reported device stability values listed in Table S2 (Supporting Information). Overall, regardless of dampness exposure and MPP tracking under light soaking, the strain-affected WBG PSCs degraded much more seriously, revealing the importance of strain control for WBG perovskite industrialization.

Effect of Strain on Phase Stability
The photo-induced phase segregation of two WBG perovskites was carefully investigated and clarified via PL measurement in Figure 5. [56][57][58] Here we utilized a Xenon lamp including UV region with an intensity of 2-sun to accelerate the segregation process. Light irradiated from the upper surface of perovskite to exclude the effect of tin oxide photocatalysis. For the 1.67 eV perovskite, obvious phase segregation occurred in the strainaffected perovskite, as shown in Figure 5a,c, with prolonged light soaking. A second PL peak was observed at ≈800 nm under tensile stress, whereas PL peak underwent a redshift under compressive stress, possibly resulting in poor MPP stability. A slight broadening of the PL peak of the strain-free perovskite in Figure 5b suggests that no phase segregation occurred. The photo-stable perovskite film supports devices with enhanced MPP stability. For 1.77 eV perovskite from Figure 5d-f, the overall stability is poorer than that of the 1.67 eV perovskite due to more Br and Cs addition. Though phase segregation occurs in all films, it is less serious for the strain-free perovskite, which breeds a small peak at ≈775 nm. Moreover, the PL emission peak at ≈625 nm is present both for the strained 1.77 eV perovskite under light soaking, which is attributed to a signal of a new-born perovskite with higher Br content or even a nearly all-inorganic perovskite phase since CsI/CsBr was first introduced in the first step deposition with PbI 2 /PbBr 2 . The strain in perovskite can decrease the formation energy of defects and activation energy for ion migration. [59,60] Thus, severer phase segregation can occur in the strained perovskites. The results also agree with the results of the stability tests in Figure 4g,h and Figure S38 (Supporting Information).

Application Expansion in Pin and Tandem Solar Cells
To further understand the universality of the process and concept, we fabricated pin solar cells with an ITO/NiO/perovskite/C60/BCP/Ag structure (Figure 6a). Based on the similar thermal expansion coefficient of SnO 2 (1.17 × 10 −5 K −1 ) and NiO (1.2 × 10 −5 K −1 ), 1.67 eV perovskite with a default 25% DMSO content was not affected by strain. [61,62]   the strained devices (Figure 6b,c). According to the IEC 61215 standard, we first tracked the damp-heat (80% RH, 80 °C) tests, as shown in Figure 6e, using sealed devices. The devices with 25% DMSO retained over 80% of their initial efficiency after 1000 h; however, they degraded rapidly with 10% or 40% DMSO. The devices exhibited less than 2% PCE loss after 80 h of UV exposure with the power of 60 kWh cm −2 , as shown in Figure 6f. This is mainly due to the UV-insensitive of NiO layer. Finally, we performed the MPP tracking test for the devices (temperatures: 55 °C ± 5 °C) under 1-sun illumination for 500 h without UV-cut. The devices with 25% DMSO can achieve over 82% of their initial efficiency, showing much better stability than the strained devices.
Finally, we fabricated semitransparent PSCs and perovskite/ Si tandems with deliberately thickened 15-nm BCP to buffer the sputtering damage of 100-nm IZO deposition. The semitransparent device with a MgF 2 antireflective layer can deliver an efficiency of 18.36% (Reverse Scan) and an integrated current density of 19.72 mA cm −2 from EQE curves as shown in Figure S39 (Supporting Information). With the champion pin configuration adjusted to the Si heterojunction half cell (Figure 6h), the following values were determined from the reverse J-V curve in Figure 6i: V OC of 1.90 V, J SC of 19.48 mA cm −2 , and FF of 76.42%. We achieved an overall efficiency of 28.35% with a minor hysteresis. The integrated J SC value of the perovskite sub cell was 19.43 mA cm −2 , which agrees well with the J SC value of 19.28 mA cm −2 for Si sub-cell (Figure 6j). We counted the performance of 30 tandem devices and obtained an average efficiency of ≈26.8% as shown in Figure S40a (Supporting Information). The champion device exhibited great stability of 10 min steady-state power point test and retained over 97% of its initial efficiency after 210 h under MPP tracking in N 2 ( Figure S40b,c, Supporting Information). The direct integration in tandem demonstrates a successful strain-free WBG perovskite application.

Conclusion
We have summarized a simple preparation method for efficient and stable WBG PSCs via strain control. The photoinduced phase segregation in WBG perovskite is correlated with the intrinsic stability of strain-affected perovskite. Two PSCs with the bandgap of 1.67 and 1.77 eV commonly used in tandem solar cells were realized with record efficiencies of 22.28% and 20.45% for nip configuration to date. Furthermore, a state-ofthe-art efficiency of 22.31% for 1.67 eV pin solar cells together with a competitive efficiency over 28% for the perovskite/Si tandem solar cells were obtained. The residual stress in perovskite can cause serious photoinduced phase segregation, reducing the stability of devices under continuous light soaking. Both the target 1.67 and 1.77 eV nip WBG PSCs showed promising results, retaining over 90% and 80% of their initial PCEs after 700 h MPP tracking respectively. Further, the 1.67 eV pin WBG PSCs with restrained strain presented great damp-heat, pre-UV aging, and MPP tracking stability tests with IEC 61215 standards. Strain effects on the efficiency and stability of WBG perovskites are critical, standing to be a key to facilitating the breakthrough of perovskite-based tandem solar cells.
Fabrication of Pin WBG Perovskite Solar Cells: The tin-doped indium oxide (ITO) glass was successively cleaned by sonication with deionized water, ethanol, acetone, and isopropanol as agents followed by UV-ozone for 20 min. NiO X nanoparticle (20 mg mL −1 in deionized water) was spincoated on the ITO substrates at 4000 rpm for 30 s with post-annealed at 100 °C for 10 min in air. The preparation of the WBG perovskite films was the same as that in conventional nip structure. Then, 25 nm C60 and 5 nm BCP were thermally evaporated at a vacuum of < 3 × 10 −4 Pa. The devices were completed with 100 nm Ag as electrodes.
Fabrication of Perovskite/Si Solar Cells: Double-side-polished phosphorus-doped Czochralski silicon wafers were fabricated with a resistivity of 1-5 Ω cm −1 and a thickness of ≈280 µm. The texturing process was conducted on one side of the wafer in an alkaline solution to obtain the random pyramids. After RCA cleaning process, a buffered oxide etching solution was used to remove the resulting oxide layer. Afterward, the intrinsic, n-and p-type a-Si:H films were deposited by plasma-enhanced CVD with thicknesses of 8, 8, and 10 nm, respectively. A 40-nm-thick ITO layer was sputtered on the textured rear side with a 90/10 In 2 O 3 /SnO 2 target followed by a 200-nm-thick Ag layer. The ITO recombination layer (40 nm) was sputtered on the front polished side of the wafers. The Si bottom cells were next annealed at 200 °C for 10 min to recover the sputtering damage. For the fabrication of the perovskite top cells, the same NiO X and WBG perovskite deposition as described above was implemented on the wafer. Subsequently, C60 (15 nm) was thermally evaporated at a rate of 0.1 Å s −1 onto the perovskite substrates. The thickness of the evaporated BCP was optimized up to 15 nm to compensate for the loss of the IZO sputtering bombardment. One hundred nanometer IZO (50 Ω sq −1 ) was sputtered on the BCP layer with a 3 inch target (In 2 O 3 /ZnO, 90%/10%) at a RF power of 50 W. Finally, Ag front electrodes (200 nm) and MgF X antireflection films (100 nm) were formed by thermally evaporating at a rate of 0.8 Å s −1 .
Materials Characterizations: Surface morphology and cross-sectional image of perovskite and the whole device were taken by a Field-Emission Scanning Electron Microscope (FESEM, S4800, Hitachi). Conventional XRD patterns of perovskite were tested by a D8 Discover (Bruker) with a Cu Kαradiation. While strain characterizations of the films were tested by X-ray diffractometer (Bruker D8, ADVANCE) with a stress attachment and a Raman spectrometer (Horiba Evolution). The bandgap of perovskite was determined using a UV-vis spectrometer (U-4100, Hitachi Limited). The steady state PL measurements (FLS980) were performed with a continuous Xenon light source (spot area 0.28 cm 2 ). The TRPL results were tested using a 405 nm excitation laser (repetition rate of 10 kHz, pulsed duration 60 ps, spot area 0.5 mm 2, and a photon flux of ≈1 µJ cm −2 ). From these data, we extract the normalized PL intensity dividing the spectrum by the maximum peak value (unit counts). The PL was collected with High-Gain Photomultiplier Tube Detectors (Edinburgh). The step width was set up to 10 nm s −1 . Load-Penetration depth curves for perovskites were obtained using a Nano-Indentation (Bruker Hysitron TI980). Here we conducted all measurements on samples with the structure of glass/patterned ITO/SnO 2 /perovskite.
Device Characterizations: Current density-voltage (J-V) curves were acquired under a simulated AM 1.5G illumination (94022A, Newport) using a Keithley 2400 in air, calibrated by a standard Si solar cell (PVM937, Newport) with KG5 filter. A 150 W xenon lamp (UXL-150MO, Ushio, Japan) was applied as a light source. All J-V characteristics were measured with a mask (0.056 cm 2 ). External quantum efficiency (EQE) spectra of devices were tested using a QE system (Model QEX10, PV Measurements, Inc.) Device Stability: For damp test, all the devices were stored under dark for the first 700 h in a nitrogen glove box and the next 3000 h in the ambient. Measure the efficiency every once in a while. For MPP test, each device with a considerable PCE (19-21% for 1.67 eV PSCs, 18-19%for 1.77 eV PSCs) was output under the maximum power point in a nitrogen glove box using a Keithley 2400 Source Measure Unit with a solar simulator (Enli Technology Co., Ltd., SS-F5-3A) without temperature control (constant temperature maintained at 55 ± 5 °C) nor any filter. For damp-heat stability tests, all the devices were encapsulated and placed in constant temperature and humidity test chamber (DB-TH, DANBLE). For pre-UV aging tests, the sealed devices were aged using a high-pressure mercury lamp (MUA-165, Cerma Precision, Japan).
Strain Calculation: For CTE mismatch model, the stress can be quantified as follows: where σ is the predicted stress due to the thermal expansion mismatch, E T is Young's modulus of the target film, v P is the Poisson's ratio of the target film, α S and α T are the respective thermal expansion of the substrate and target films, and ΔT is the temperature gradient during cooling from the annealing temperature of the target film to room temperature. And the strain of the target film ε is obtained by dividing σ by E T .
The measured strain can be obtained using classical sin 2 ψ equation where ψ is the crystal plane azimuth angle shown as follows: where θ 0 is the diffraction peak for stress free perovskite (hkl) crystal plane and θψ is the diffraction peak for the actual perovskite films, K is the simplified constant. For continuous, homogeneous isotropic materials, 2θ ψ ∼ sin 2 ψ is a strictly linear relationship with a slope, once the slop is determined, the related stain ε can be obtained. Considering the d spacing, the strain ε can be defined as follows: where d is the measured d spacing, d 0 is the d spacing of strain-free films.

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