Ethylenediamine Vapors‐Assisted Surface Passivation of Perovskite Films for Efficient Inverted Solar Cells

Defects present at the surface or within the bulk of halide perovskites act as a barrier to charge transfer/transport, induce nonradiative recombination thereby limit open‐circuit voltage (VOC), and accelerate degradation in the perovskite solar cells (PSCs). Passivation of these defects at surfaces, interfaces, and grain boundaries to suppress the charge recombination is therefore imperative to improving photovoltaic performance in the PSCs. Herein, a facile posttreatment of perovskite surface by ethylenediamine (EDA) via mixed solvent vapor annealing method is reported. The results show that only a trace amount of EDA causes significant suppression of nonradiative recombination leading to over 100 mV increased VOC and ≈22% improvement in power conversion efficiency (PCE) of the inverted PSCs. The key reasons for this improvement are an upward shift in the Fermi energy level, reduced lattice strain and Urbach energy, and reduction in nonradiative recombination upon EDA passivation. These lead to a PCE exceeding 20% up from 16% for a nonpassivated film. The unencapsulated EDA‐modified PSCs also demonstrate an improved shelf‐life and retain 87% of the initial PCE after 850 h.

to passivate the bulk and interfacial defects. The inclusion of amphiphilic t-butyl cations created 2D-3D heterojunction (through diffusion from interface to bulk) that improved the hole extraction and extended stability with high FF (84.1%) and V OC (1. 19 V). [13] In another work, Zang et al. overcome interfacial defects and energy barrier by incorporating a series of guanidinium salts with various anions at perovskite/ETL interface and achieved an improved PCE with high FF (82.7%). [14] Various other strategies have also been proposed to overcome the V OC and FF loss in perovskite solar cells (PSCs) such as surface passivation, compositional engineering, additive engineering, solvent engineering, and interface passivation via insertion of interlayers. [7,15] Among these, surface modification approach is an attractive strategy that has resulted in improved device performance and stability. While the former is mainly due to defect passivation, the latter is because the thin modifier layer has a superior resilience to the environmental effects. [16] The surface engineering strategies often include physical or chemical methods of passivation with their distinct benefits including separation of defects from minority charge carriers and reduction of defect states in the perovskite films, respectively. [17] Snaith et al. introduced a Lewis base passivation concept for perovskite layer by using the thiophene, pyridine, and iodopentafluorobenzene through supramolecular halogen bonding that effectively reduced the nonradiative recombination. [18] Similarly, the use of long-chain hydrophobic organic molecules resulted in improved performance and stability of PSCs. [19] Alkylamines, when applied to the surface or added into the precursor solution, have also shown to improve the V OC in a variety of perovskites such as MAPbI 3 , FAPbI 3 , and MA-, FA-, Sn-containing mixed perovskites. [20] Bakr and co-workers studied the influence of butylamine, phenethylamine, octylamine, and oleyl amine added in the precursor solution of triple cation perovskite and found that the oleyl amine (large-chain alkylamine among the series investigated in their work, C 18 H 35 NH 2 ) efficiently reduced the trap states and improve the carrier mobilities that translated into an enhanced PCE. [21] Huang et al. [8] also demonstrated an enhanced performance in the PSCs from 18.3 to 20.0% when short-chain bilateral propylamine added into the MAPbI 3 precursor solution. Recently, ethylenediamine (EDA) and its derivatives including ethanolamine and EDA chlorides have been reported to suppress the defect states in all inorganic perovskite, [22] Sn-Pb perovskite [20d,23] and triple cation perovskite, respectively. [24] It, however, remain unclear as to how the EDA contributes to the morphology, defect passivation, and optoelectronic properties of perovskite films. A detailed study on the role of the EDA toward morphology, surface passivation mechanism, defects and recombination in the perovskite films, energetic alignment, and charge extraction at the device interfaces is highly desired.
In this work, the role of EDA on the structural, morphological, and optoelectronic properties of MAPb(I/Cl) 3 (hereafter referred as MAPbI 3 ) films for inverted PSCs is systematically explored. Our approach includes a facile and highly reproducible strategy of posttreatment of MAPbI 3 surface in the presence of EDA vapors (e.g., a mixed solvent vapor annealing). It was noted that the terminal diamine groups from EDA passivate the defects and dangling bonds due to undercoordinated Pb 2þ sites and iodine vacancies, thus reducing the trap state density in the EDA-passivated films. The EDA passivation also enhances crystallinity of perovskite film and leads to a densely packed film morphology featuring larger grains than the nonpassivated counterpart. Kelvin probe force microscopy (KPFM) analysis reveals an upward shift of the Fermi energy level (E f ) suggesting a more electronegative behavior, while photoluminescence (PL) and space-charge-limited current (SCLC) measurements show a reduced defect density upon EDA passivation. All these contributed significantly to the improved photovoltaic performance in p-i-n PSC architecture with EDA-passivated PSC showing over 100 mV increase in the V OC and a relative 22% improvement in the PCE. The origin of the improved performance has been investigated comprehensively through an extensive characterization of perovskite films and devices.

Fabrication of Perovskite Films and Surface Passivation
The perovskite layer was deposited via a single-step spin-coating method and crystalized via a vacuum-assisted annealing, as shown in Figure 1. The EDA passivation was introduced via solvent vapor annealing followed by annealing at 90°C for 30 min. Herein, the EDA solution (used for solvent-vapor annealing) is made from various EDA concentrations (0.02 to 0.1 wt%), dissolved in isopropyl alcohol (IPA). The schematic illustration of the solvent vapor setup and device architecture adopted in this study is shown in Figure 1 and further details are presented in the ESI.

Morphology, Electronic Structure, and Defects
The structure, morphology, and optoelectronic properties of the perovskite films were first studied to probe the effect of EDA passivation. Only the optimal concentration (0.06 wt%) of EDA (taken from the device measurement data and steady state photoluminescence (PL) quenching, Figure S1, Supporting Information) was chosen for the detailed study. The surface morphology and roughness of the films with and without EDA were examined by atomic force microscope (AFM, Figure S2 and S3, Supporting Information) and scanning electron microscope (SEM, Figure 2 and S4-S8, Supporting Information). All perovskite films show a densely packed morphology. The EDA passivation, however, leads to an increase in the average grain size, which systematically increases with increasing the EDA concentration to 0.06% and then drops when the EDA concentration is 0.08% ( Figure 2f ) and higher. The increase in the grain size can be attributed to the bridging effect of the EDA molecule due to its terminal amine groups (terminal nitrogen having lone pairs). Such an effect is previously reported for alkylamines and ethane-1,2-diamine-passivated perovskite. [8,20c] The EDA as a Lewis base interacts with the surface and grain boundaries thus promoting the growth of secondary grains [25] and passivates the surface defects via filling the iodine vacancies. The secondary grains result in preferential growth of crystals, which is reflected from a superior crystallinity of EDA-treated samples than the pristine counterpart ( Figure S9, Supporting Information). EDA also has the tendency to dissolve the perovskite grains www.advancedsciencenews.com www.solar-rrl.com  [8,20d] while the values for perovskite (with and without EDA passivation) are taken from experimental measurements (shown below). The dashed line in the perovskite energetic shows the position of Fermi energy, which is calculated from KPFM measurement. c) Device architectures of reference and EDA-passivated PSCs in this study. www.advancedsciencenews.com www.solar-rrl.com through its bleaching effect (strong basic character being a Lewis base). [20c] The bleaching effect of EDA is favorable until an optimal concentration of 0.06%; a higher EDA concentration of 0.08% and onward adversely effects morphology of the perovskite films. This bleaching effect for nonoptimal EDA concentration restricts the average grain size and might lead to a higher defect density; the latter is evident from the relatively poorer performance of the PSCs employing 0.08-0.1% EDA-passivated perovskite films. Notably, the EDA treatment and coarsening of grains are also visible from the SEM images shown in Figure 2g,h clearly demonstrating that EDA interacts with the perovskite surfaces and lead to a rearrangement of surface features. The 0.06% EDA-passivated films also show a smaller root mean square roughness of 13.7 nm than that of the pristine perovskite films (14.9 nm, Figure S2 and S3, Supporting Information).
To investigate the effect of EDA passivation on the crystallization of the MAPbI 3 perovskite films, the X-ray diffraction analysis (XRD, Figure S9, Supporting Information) was carried out in ambient conditions. The main diffraction peaks at 2θ values of 14.95°, 28.45°, and 31.88°, which correspond to (110), (220), and (310) planes, do not show any shift for the pristine-and EDA-treated samples. The intensity of these peaks, however, increases for the 0.06% EDA sample suggesting a higher crystallinity. The full-width at half-maximum (FWHM) also decreases from 0.18 to 0.15 upon EDA passivation, which suggests a higher crystallinity and an increase in the grain size. The latter affirms the observations from the SEM analysis ( Figure 2). Additionally, a reduction in micro strain is also observed for the EDA-modified perovskite film (ε = 5.29 Â 10 À3 ), which suggests a superior crystallinity than the reference perovskite film (ε = 6.51 Â 10 À3 , see the detailed calculations in the Figure S9 and Table S1, Supporting Information). [26] This can be attributed to the passivation of dangling bonds arising from I À vacancies and undercoordinated Pb 2þ sites leaving less strain on the grains.
No notable change was observed in the bandgap of both films. The optical absorbance and transmittance spectra ( Figure S10 Figure S12, Supporting Information. g-i) High-resolution XPS spectra of g) Pb 4f, h) I 3d and i) N 1s core levels for 0% EDA (lower panel) and 0.06% EDA (upper panel) samples.
www.advancedsciencenews.com www.solar-rrl.com perovskite films, respectively. The values of highest occupied molecular orbital (HOMO) were estimated from photoelectron spectroscopy in air (PESA) measurements to be 5.58 and 5.30 eV for 0.06% and 0% EDA, respectively (Figure 3b,c). The lowest unoccupied molecular orbital (LUMO) was estimated from the difference of the above values and is shown in the energy diagram ( Figure 1) where a lower lying conduction and valance bands for the EDA-modified samples were noted. These results are in agreement with the past reports where EDA passivation in Pb-Sn perovskites shows a similar trend in energetics due to a decrease in the self-doping effect caused by a reduction in defects density. [20d] We propose that the nitrogen-rich surface of EDA acts as a source of free charges by donating lone pair of electrons to the perovskite, which can uplift the Fermi energy level (E F ) as well as can improve the stability of the perovskite film. The surface potential of reference and modified films was measured from KPFM in nitrogen environment, which is used to calculate the E F values (Figure 3d,e, and S2, S3 and S11, Supporting Information). The EDA-passivated film shows an upward shift in the Fermi energy suggesting a more negative character of the resultant perovskite. The shift in the E F can be attributed to electron-donating amine moieties in the EDA molecule, which changes the surface properties of the perovskite. The improved crystallinity and a smaller defect density upon EDA passivation are also evident by comparing the Urbach energy (E U ) of both perovskite films (Figure 3f ). Generally, a larger E U value indicates a higher trap states density in perovskite films. [27] The E U value for the 0.06% EDA sample is smaller (37 meV) than a reference MAPbI 3 (43 meV), indicating a lesser defect density in the former. To understand the possible interaction between EDA molecules and MAPbI 3 , X-ray photoelectron spectroscopy (XPS) measurements were carried out (Figure 3g-i). High-resolution XPS spectra of Pb 4f core level related to Pb 2þ show a peak shift toward lower binding energy (BE) upon EDA passivation. Similarly, a shift in the BE is also observed for I 3d (619.79-619.54 eV) and N 1s (402.89-402.51 eV) core levels (Figure 3h,i). The shift in the BE can be attributed to the modified chemical environment around the [PbI 6 ] 4À octahedra when EDA molecule having two terminal nitrogen atoms with lone pair of electrons interacts with undercoordinated Pb 2þ cation, which is also evident from the corresponding nonidentical work function of both films (Figure 1b). A shift in the BE of the Pb 2þ cation to lower values has been attributed to a decrease in electron affinity of Pb 2þ cation upon electron donation from the passivating molecule. [28] This affirms our hypothesis that the EDA interacts with undercoordinated Pb 2þ cation by offering its lone pair to fill the I À vacancy. EDA, a bidentate having nitrogen in its structure, is more electronegative and can also replace nearby I À anions from the [PbI 6 ] 4À octahedra. As a result, the number of undercoordinated Pb 2þ cations is suppressed, which offer overall less electron affinity and hence appeared at relatively lower binding energy. This is also confirmed by comparing I/Pb ratio in perovskite films, which is 4.4 and 4.0 for reference and EDA-passivated perovskite film, respectively, suggesting a lesser undercoordinated Pb 2þ sites in the latter. These results affirm the finding that a higher I/Pb ratio is a measure of a higher number of undercoordinated Pb 2þ sites, which is in agreement with the previously reported results where the smaller Br/Pb ratio was found suitable for lesser undercoordinated Pb 2þ sites. [29] The XPS spectra also show 14% increase in the peak intensity corresponding to the N 1s core level upon EDA passivation together with the emergence of two new peaks at 399.81 and 398.11 eV, which can be linked to the free-NH 2 and Pb coordinated -NH 2 amine group from EDA. A qualitative analysis of the various peaks shows that the relative intensity of I 3d and Pb 4f peaks decreases by 12% and 6%, respectively, which together with the 14% increase in the N1s peak confirms the presence of EDA at the surface of perovskite film.
To further probe the defect and recombination in both perovskite films, steady-state photoluminescence (SSPL) spectra and time-resolved photoluminescence (TRPL) transients of both films (Figure 4a,b) were recorded. The SSPL spectrum of the EDApassivated film on glass (nonquenching substrates) shows a red shift of around 3 nm and a doubled PL emission. Such a strong increase in the PL emission from 0.06% EDA sample despite its identical absorbance to that of the reference perovskite ( Figure S10, Supporting Information) suggests a higher radiative yield due to reduced nonradiative recombination sites. The higher PL intensity of the EDA-passivated perovskite is also backed from its higher crystallinity, as can be seen from the XRD measurements ( Figure S9, Supporting Information). [30] This is further verified from the TRPL curves of the two samples: The 0.06% EDA sample shows a smaller drop in the PL intensity during first few ns, which is another indication of a smaller trap density in the film, [31] which in this work is obtained via EDA passivation.
The trap density was calculated in the two perovskite films via SCLC measurements. The devices were made in a hole only configuration to best match with the PSCs in this work. For reliability, the mean SCLC values ( Figure S13, Supporting Information) and the best representation of both perovskites are provided in Figure 4e,f. Clearly, the EDA passivation leads to around four-time reduced trap density (N T = 8.8 Â 10 16 cm À3 ) as compared to the reference perovskite films (0% EDA; N T = 29.5 Â 10 16 cm À3 ). The trap density is calculated from the trap-filled region in the current-voltage (JV) curve ( J ∝ V 2 , n ≥ 2), which corresponds to a region where traps are filled with charge carriers (see a detailed description of SCLC method in the supporting information).
One must note that the PL of the perovskite films on glass substrate might not be a true representation of the recombination in a full device, which involves deposition of perovskite film over an HTL (PTAA in our case). This is because the growth and properties of perovskites show substrate-dependent optoelectronic properties [32] and also the recombination kinetics of perovskites on quenching and nonquenching substrates might largely differ due to different interfacial recombination. [33] For this reason, the SSPL and TRPL on PTAA-coated ITO substrates were also measured, which unlike the glass substrates, allow charge extraction from the perovskite film. It is noted that the PL intensity of the 0.06% EDA sample shows a strong reduction in the PL intensity (63% lower than the reference sample, Figure 4c,d), which can be attributed to quenching of charge carriers. The SSPL spectra of the entire EDA doping concentrations (0, 0.02, 0.04, 0.06, and 0.08% in Figure S1, Supporting Information) also confirm improved charge extraction until 0.06% EDA doping and a further increase in the PL intensity (a reduced charge extraction) in the case of 0.08% EDA doping.
The TRPL transient curves of both perovskites on PTAA-coated ITO substrates further confirm a faster charge extraction in the EDA-doped perovskite film (the nonnormalized TRPL data show a higher PL count in the EDA-passivated film indicating a higher radiative recombination). Fitting the TRPL transients with a biexponential decay function yielded a shorter average lifetime of 4.64 ns for the EDA-doped perovskite, which is around 3-4 times shorter than undoped counterpart (17.67 ns). A detailed analysis of the fast (τ 1 ) and slow (τ 2 ) components of the TRPL transients corresponding to nonradiative and radiative recombination, respectively, is shown in Table S2, Supporting Information. We would like to emphasize that the term charge carrier lifetime is often confused with charge extraction time and that a difference between these two terms should be shown by considering different experimental designs (PL studies on both quenching and nonquenching substrates). In this work, when perovskite film is excited (from the PTAA/perovskite interface side) using a pulsed laser, an initial population of charge carriers (n) is reached (showing the initial maximum PL intensity) immediately. Switching off the pulse leads to depopulation of charge carrier through various recombination mechanisms. Charges (hole in this case) are extracted to charge selective layers (PTAA), which is reflected in the initial drop in the PL intensity. Since these charges cannot be extracted to an external circuit (as the device is incomplete), these holes diffuse through the PTAA film until they find electron/trap states and recombine (in the case, this recombination is radiative, this can be seen from the tail of the TRPL or the slower component of recombination). The faster drop in the case of EDA-doped perovskite thus suggests a more efficient charge extraction due to the improved energy alignment and less trap states, which is evidenced from Figure S1, Supporting Information.  www.advancedsciencenews.com www.solar-rrl.com

Photovoltaic Performance and Device Physics of the Perovskite Solar Cells
The photovoltaic performance of PSCs employing nonpassivated and EDA-passivated perovskite films is shown in Figure 5 as well as in Table S3 and Figure S14 and S15, Supporting Information. We measured performance of devices as a function of a large variation in the EDA concentration (0.005% to 0.2%) to demonstrate the effect of insufficient and excess passivation, respectively. The current-voltage ( J-V ) curves of PSCs with various EDA concentration are given in Figure 5a, and the photovoltaic (PV) parameters are plotted in supporting information ( Figure S14 and S15, Supporting Information). The PV performance increases with increasing EDA concentration up to 0.06% and decreases afterward. The most notable is the trend in the V OC ; an increase of around 106 mV is noted between reference (0%) and 0.06% EDA containing PSCs. The V OC gradually drops with further increasing the EDA concentration. This, together www.advancedsciencenews.com www.solar-rrl.com with the drop in the FF and the J SC , could be attributed to the poor film morphology and a poor charge extraction once the EDA concentration exceeds 0.06%. Figure 5b compares the PV performance of the nonpassivated and optimized EDA (0.06%) devices. The nonpassivated device showed a PCE of 15.9% with V OC of 0.99 V, J SC of 20.4 mA cm À2 , and FF of 73.5%. The optimal EDA concentration (0.06%) owing to the larger grain size, a higher crystallinity, and improved charge carrier lifetime showed 106 mV higher V OC (1.1 V) accompanied by an improved J SC of 24.0 mA cm À2 and FF of 79.9%, thus leading to a PCE exceeding 20%. We attribute the increase in the J SC (as well as other PV parameters) to the following: 1) the reduction of various recombination centers at the perovskite surface and ETL/perovskite interface, as evident from Figure 4; 2) a more favorable energetic alignment upon EDA passivation (evident form a faster charge extraction, see Figure 4 and 6); and 3) the improved crystallinity and a higher grain size in the EDA-passivated films. To validate the J SC in the devices, external quantum efficiency (EQE) of the bestperforming reference and 0.06% EDA-containing PSCs (Figure 5c) are shown. While the EQE onset remains constant for both the devices, a significantly improved charge collection is evident for EDA-passivated PSC in the entire visible spectrum. Integrating the EQE spectrum with the photon flux at standard measurement conditions yields J SC values of 20.1 and 23.6 mA cm À2 for reference and 0.06% EDA-passivated PSCs, respectively, thus affirming a higher charge collection in the latter. The increase in the V OC and the FF suggests a suppressed nonradiative recombination and efficient charge extraction in the EDA-modified perovskite films, which is established from TRPL measurements and from leakage current ( J sh ) measurements. The J sh in solar cells is undesirable current that flows opposite to the photogenerated current and hence limits the J SC . [34] A quantification of the J sh through the device can be used to probe the nonradiative recombination of charge carriers. Figure 5d shows that the EDA-passivated PSC shows a lower J sh in comparison to the nonpassivated PSC. This can be understood by comparing the shunt resistance (R sh ), which is related to the J sh via the relation (J sh ¼ VÀJR s R sh ; here R s is series resistance). [34b] The R sh of the EDA-passivated cell exceeds 3000 Ω cm 2 , up from 170 Ω cm 2 for a reference cell, thus affirming a suppressed nonradiative recombination upon EDA passivation.
The maximum power point (MPP) tracking measurements were performed to show reliable efficiency of our devices, following a protocol developed by Zimmerman et al. [35] Figure 5e shows the stabilized PCE of EDA-passivated and nonpassivated PSCs, measured at the MPP under AM 1.5G illumination. The EDA-modified champion PSC shows a stabilized PCE of 20.1%, Figure 6. Carrier recombination dynamics, ideality factor, and LID measurements. a) Charge extraction lifetime measured by transient photocurrent (TPC) for 0% and 0.06% EDA devices. b-d) J SC , V OC , and FF dependence as a function of light intensity for 0% and 0.06% EDA devices.
www.advancedsciencenews.com www.solar-rrl.com which is identical to the PCE calculated from the J-V curves (PCE 20.2%). A statistical analysis of over 40 selected devices with EDA (0.06%) and 36 reference devices is also provided ( Figure S15, Supporting Information) that confirms the trends as shown in Figure 5b. The reproducibility can also be confirmed from the data of 60 random devices of each, with and without EDA, from various batches shown in the bar graph (Figure 5f ). Generally, a higher PCE is evident for 0.06% EDA-containing PSCs affirming a higher reproducibility. EDA passivation leads to an improved average PCE of 19.4%, which is a 23% improvement as compared to that of the reference PSC with 0% EDA. The suppression of nonradiative recombination upon EDA passivation is further evidenced through transient photocurrent (TPC) measurements and light intensity dependent (LID) J-V measurements. TPC measurements can provide information about carrier recombination lifetime, which we measured for representative devices of both the reference and EDA-modified PSCs. The EDA-passivated PSC showed a shorter charge extraction time (848 μs) than the reference PSC (0% EDA, 1922 μs), as shown in Figure 6a. The shorter charge extraction time suggests an efficient charge extraction from perovskite absorber layer to PTAA HTL as well as reduced charge recombination for EDA-modified devices than MAPbI 3 only devices. This observation is consistent with the higher J SC and FF values for the EDA-modified devices. The probable explanation is that the EDA eliminates the undercoordinated Pb sites, which would act as nonradiative recombination centers. [36] The LID measurements are helpful in identifying whether the recombination is related to the interfacial, Shockley-Read-Hall (SRH) bulk recombination or the recombination at metal contact. [37] This can be done by calculating ideality factor (n id ) from LID measurements. The light intensity dependence of the PV parameters, i.e., J SC , V OC , and FF are shown in Figure 6b Table S(4 and 5), Supporting Information, respectively. The linear relationship of J SC with the light intensity demonstrates absence of a significant energy barrier in the devices. The J SC dependence on light intensity is usually monitored by a fitting factor α, which is associated to bimolecular radiative recombination. [38] Apparently, a linear correlation is observed between light intensity and J SC for all devices with the values 0.99 and 1.05 for the EDA-modified and reference devices, respectively, suggesting insignificant contribution of bimolecular recombination. The relationship between V OC and light intensity is shown in Figure 6c. It is established that trap-assisted recombination can be probed by monitoring slop deviation from (K B T/q). [16] The value of the slope is calculated to be 1.28 and 1.87 K B T/q for EDA modified and reference device, respectively. The smaller slope value for EDA sample clearly demonstrates that the EDA-modified PSC is characterized by fewer trap states advocating effective surface passivation.
Notable is the FF trend with varying light illumination in both samples. The dependence of the FF on light intensity, particularly at low light intensity regime, is helpful to probing the trap-assisted recombination in PSCs. [8] Figure 6d shows the light intensity dependence of the FF of both devices. While the FF in the reference PSC drops as the light intensity decreases from 1 to 0.01 sun, the FF in the EDA-modified device remains flat in the entire measurement range. Glowienka and Galagan suggested that the decrease in the FF with reducing light intensity is caused by the defects within the absorber layer and at its interfaces. [39] This suggests a predominant trap-assisted nonradiative recombination owing to surface trap states in the reference PSC and that the EDA effectively passivated the surface traps.

Operational Stability of the Perovskite Solar Cells
In order to explore the effect of the varying defect density on the operational stability of devices, the shelf-life stability of the nonencapsulated PSCs for around 850 h (Figure 7a,b) was measured. Clearly, the EDA-modified device showed significantly improved stability by retaining 88% of its initial PCE after 850 h, whereas the PCE of the reference device dropped by 66% within 500 h. The faster PCE drop in the reference PSC is caused mainly due to the decrease in the FF and V OC , while the J SC remained nearly constant throughout the measurement. It is proposed that the EDA molecules passivate undercoordinated ions at surfaces and grain boundaries of perovskite crystals and hence reduce the www.advancedsciencenews.com www.solar-rrl.com defect induced degradation, which is evident from long-term stability of the PSCs. To understand the difference in the operational stability of both devices, the hysteresis index (HI) over the period of their shelf-life testing was also compared. The HI increases drastically to a value close to 0.4 for the reference device and remained below 0.05 for EDA-modified devices. The HI strongly depends on both the internal (material related properties) and external factors. Care must be taken to exclude any variation due to external factors such as measurements and atmospheric conditions. [40] In our case, all the external factors, e.g., scan speed, scan range, prebiasing, measurement delay, and atmospheric conditions, remain the same for both devices, and therefore, the change in the HI over time can be linked to a change in material properties, e.g., ionic mobility, defects at the grain boundaries and at the heterointerfaces, delayed charge carrier transportation, and extraction and finally the recombination (surface and bulk), etc. [40a] The trends of HI indicate that the nonpassivated perovskite film's surface in the reference PSC deteriorates over time due to a higher trap density, and therefore, recombination that limits charge extraction at the MAPbI 3 /ETL interfaces reduces the PCE. This is not the case for the EDA-modified PSCs where an effective passivation of the trap states, both at the surface and at the grain boundaries, led to an elongated operational stability for over 850 h. The observation related to HI is affirmed by the finding of the PL, TRPL, LID, TPC, and JV measurements.

Conclusions
In this work, the importance of surface defect passivation in the archetypical MAPbI 3 perovskite is demonstrated. The defect passivation is caried out via exposing an already annealed perovskite film to EDA vapors at room temperature. The EDA vapors lead to the growth of secondary grains, which together with coarsening of grain boundaries result in an enhanced grain size. SEM and XPS experiments confirm the presence of EDA molecule on the surface of perovskite film, whereas KPFM measurements show an upward shift in the Fermi energy suggesting a more electronegative character of the resultant perovskite. The shift in the E F can be attributed to electron-donating amine moiety in the EDA molecule, which changes the surface properties of the perovskite.
Space charge limited current measurements show a four times reduced trap density (N T = 8.8 Â 10 16 cm À3 ) in the EDA-passivated film as compared to the reference perovskite films (0% EDA; N T = 29.5 Â 10 16 cm À3 ). Steady-state and time-resolved photoluminescence measurements also evidence an enhanced radiative yield and an enhanced carrier lifetime upon EDA passivation. The reduction in defect density is also evidenced from micro-strain (ε) calculations (the value of ε drops from 6.61 Â 10 À3 to 5.29 Â 10 À3 for the optimal EDA concentration of 0.06%).
When applied in p-i-n PSC architecture, the EDA-passivated films show nearly 100 mV enhanced V OC and a higher FF, both suggesting a reduced nonradiative recombination. The enhanced J SC upon EDA passivation is attributed to a superior charge transfer due to a more aligned energetics, which is supported by enhanced charge extraction in TRPL measurements of the perovskite film deposited on top of an HTL. The EDA-passivated devices show a PCE exceeding 20% up from 16% for a nonpassivated film and a higher shelf-life stability after 850 h.

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