Amorphous BaTiO3 Electron Transport Layer for Thermal Equilibrium‐Governed γ‐CsPbI3 Perovskite Solar Cell with High Power Conversion Efficiency of 19.96%

Compared to organic–inorganic hybrid perovskites, the cesium‐based all‐inorganic lead halide perovskite (CsPbI3) is a promising light absorber for perovskite solar cells owing to its higher resistance to thermal stress. Nonetheless, additional research is required to reduce the nonradiative recombination to realize the full potential of CsPbI3. Here, the diffusion of Cs ions participating in ion exchange is proposed to be an important factor responsible for the bulk defects in γ‐CsPbI3 perovskite. Calculations based on first‐principles density functional theory reveal that the [PbI6]4− octahedral tilt modifies the perovskite crystallographic properties in γ‐CsPbI3, leading to alterations in its bandgap and crystal strain. In addition, by substituting amorphous barium titanium oxide (a‐BaTiO3) for TiO2 as the electron transport layer, interfacial defects caused by imperfect energy levels between the electron transport layer and perovskite are reduced. High‐resolution transmission electron microscopy and electron energy loss spectroscopy demonstrate that a‐BaTiO3 forms entirely as a single phase, as opposed to Ba‐doped TiO2 hybrid nanoclusters or separate domains of TiO2 and BaTiO3 phases. Accordingly, inorganic perovskite solar cells based on the a‐BaTiO3 electron transport layer achieved a power conversion efficiency of 19.96%.


Introduction
After their first application in photovoltaic devices by Miyasaka et al., metal halide perovskite (ABX 3 ) materials became the forerunners of photovoltaic technology research. [1]The photoelectric properties, including the bandgap, carrier diffusion length, and absorption coefficient, of the PSCs are titanium oxide (TiO 2 ) and tin oxide (SnO 2 ), which were originally developed for hybrid organic-inorganic lead perovskites (HOIPs) with a lower conduction band minimum (CBM). [24,25]Consequently, imperfect band alignment inevitably exists at the interface between CsPbI 3 and the ETL, resulting in interfacial nonradiative recombination due to inefficient charge transport. [26]Accordingly, the potential loss caused by mismatching energy levels limits the PCE of CsPbI 3 PSCs.In a few studies, an interfacial layer was introduced to decrease nonradiative recombination at the electron transport interface.For instance, Ye et al. [27] introduced an inorganic layer, lithium fluoride (LiF), for blocking the electrical shunt between SnO 2 and CsPbI 3−x Br x , resulting in interfacial defect passivation and enhanced energy band alignment.Chang et al. [28] utilized Zn(C 6 F 5 ) 2 as a functional molecule to modify the misalignment of energy at the interface of CsPbI 3 and SnO 2 .To manipulate the band offsets at the interface in TiO 2 -based CsPbI 3 PSCs, some researchers modified the surface electronic state of the TiO 2 layer by employing 6,6-phenyl C 61 -butyric acid (PCBA), [29] phenyltrimethylammonium bromide (PMABr), [30] 5-amino-2,4,6-triiodoisophthalic acid (ATPA), [31] and europium acetate (EuAc 3 ), [32] which were able to mitigate the energy misalignment.However, managing the surface potential of ETLs through the deployment of binding molecules is not permanent under external stresses such as UV light, heat, and electrical bias.Additionally, the development of fundamental materials as ETL for CsPbI 3 PSCs has been overlooked.Thus, an alternative ETL with desirable characteristics such as a wide bandgap, decent light stability, and a suitable conduction band edge relative to CsPbI 3 perovskite materials could be expected to play a significant role in enhancing the performance of all-inorganic PSCs.
In this study, we demonstrated that the diffusion of Cs ions for ionic exchange during the growth of γ-CsPbI 3 films affects the crystal structure, electronic structure, morphologies, and optical properties of γ-CsPbI 3 .Their diffusion is apparently governed by the relative humidity (RH) and thermal equilibrium (TEQ) in air, which substantially affects the quality of the final film of the γ-CsPbI 3 perovskite.Strain-relieved γ-CsPbI 3 films with a bandgap of 1.71 eV are deposited by adjusting the crystallization process.Moreover, for the first time, we introduce amorphous barium titanium oxide (a-BaTiO 3 ) as a promising alternative to TiO 2 for CsPbI 3 PSCs.Regarding the energy levels, the a-BaTiO 3 exhibits a 0.19 eV band offset with CsPbI 3 (0.42 eV for TiO 2 ), and the well-aligned energy levels suppressed potential loss and interfacial trap states.The corresponding device reached the maximum open-circuit voltage (V oc ) in a matter of seconds, achieving an impressive PCE of 19.96% and steady-state output of 19.6%.

Results and Discussion
Thin CsPbI 3 perovskite films were annealed in a glove box in which the RH was precisely controlled.The corresponding final CsPbI 3 devices were characterized by sweeping the current density-voltage (J-V) curves under one-sun illumination.As depicted in Figure S1, Supporting Information, the performance of the CsPbI 3 PSCs appeared to be affected by the humidity, and the optimal performance (~16.9%) was achieved with perovskite films annealed in ambient air condition with an RH of approximately 10%.The photovoltaic parameters, such as the V oc and fill factor (FF), primarily influenced the PCE of the devices, because these parameters are related to the trap density of states in the perovskite films.To investigate the effect of humidity conditions on the formation of the CsPbI 3 crystals, high-resolution X-ray diffraction (XRD) measurements were performed on annealed CsPbI 3 film samples (Figure 1a), after which all the diffraction patterns were fitted with the split pseudo-Voigt function.The XRD patterns of all samples were characteristic of γ-CsPbI 3 perovskite, with the major diffraction peaks corresponding to the (110) and (220) planes at 14.3°and 28.8°, respectively.They were fine-tuned based on atmospheric conditions as the diffraction peak shifted to higher angles.
We also extracted the orthorhombic unit cell lattice parameters associated with the γ-phase to further understand the extent to which the changing d-spacing corresponds to the changes in the crystal structure (Figure 1b).Notably, γ-CsPbI 3 is an orthorhombic crystal system of which the lattice parameters a ≠ b ≠ c.We observed a 0.5% decrease in the volume of the unit cell due to atmospheric conditions.This observation suggests that the formation of the γ-CsPbI 3 crystal becomes more densely packed in the presence of moisture than in the absence of thereof.In addition, calculation of the crystal strain exerted on the γ-CsPbI 3 films (Figure S2, Supporting Information) [33] revealed that the strain decreased by between 10 and 20% RH (Figure 1c) and began to rise as the RH level increased.
To further investigate the morphological evolution, we analyzed the γ-CsPbI 3 films using field-emission scanning electron microscopy (FE-SEM).The varying grain sizes and morphologies of the γ-CsPbI 3 films were caused by the moisture content.Small pillar-like residues remained on the samples prepared in the presence of extremely low levels of moisture (Figure 1d,e). [34]The absence of burning and charging effects on the samples led us to believe that the residues were inorganic rather than organic, such as remaining DMAI. [34,35]Additionally, the morphology of the DMAI residues was not observed in the samples.At 10% and 20% RH, comparable dense thin films with large grain size (~400-900 nm) were observed (Figure 1f,g).By contrast, grain boundary pinholes became dominant at 30% RH (Figure 1h). Figure 1i depicts the photoluminescence (PL) optoelectronic analysis performed on the respective films to investigate their optical properties.Strong radiative recombination was observed in the 10% RH film, corroborating the low bulk defect density of the film.Interestingly, the PL peak wavelength underwent a bandgap shift, based on the atmospheric condition during thermal treatment.The calculated PL peaks of perovskite films annealed under N 2 and at RH of <1% (both denoted as inert for clarity) were 1.73 eV, which was in good agreement with the previously reported bandgap of γ-CsPbI 3 .The bandgaps of the perovskite films heated at 10 and 20% RH were calculated to be 1.71 and 1.72 eV for 30% RH, respectively.Furthermore, blue shift was detected with a low PL intensity for the N 2 and RH 1% films, but with a higher intensity than for RH 30%.The observed PL spectra indicate the improbability of shallow trap states inducing minute shifts in the bandgap.
The changes in the optical bandgap of γ-CsPbI 3 were hypothesized to be closely associated with the incorporation of the large dimethylammonium (DMA) cation into the A-site of CsPbI 3 . [36]Considering the sublimation temperature of DMA (~120 °C), we hypothesized that the changing optical properties of γ-CsPbI 3 corresponded to changes in the crystal structure caused primarily by the three-dimensional (3D) framework of the corner-sharing [PbI 6 ] 4− octahedra intercalated by the Cs cation, [37] rather than by the remaining DMA.To determine the fundamental properties of the structural effect on the physical properties of γ-CsPbI 3 perovskite, we employed first-principles density functional theory (DFT).Figure 2a depicts the DFT modeling results of the γ-CsPbI 3 structure derived from the XRD pattern.Considering the diffraction data, the calculated atomic positions revealed an adjustment of the [PbI 6 ] 4− octahedral tilt according to the unit cell volume depending on Energy Environ.Mater.2024, 7, e12625 the fabrication conditions.In particular, a decrease in the Pb-I-Pb bond angle (155.34°) was observed in the crystal structure of γ-CsPbI 3 annealed under inert conditions, resulting in low symmetry and increased tilting.Under moisture-assisted annealing conditions, the γ-CsPbI 3 crystal structure exhibited improved structural symmetry with a Pb-I-Pb bond angle of 156.29°.The bandgap of the corresponding atomic structures was computed to correlate the effect of the [PbI 6 ] 4− octahedral tilt with the band structure (Figure 2b).We accounted for the spin-orbit coupling (SOC) in the bandgap calculation by utilizing the generalized gradient approximation with the Perdew-Burke-Ernzerhof (GGA-PBE) component of the CAmbridge Serial Total Energy Package (CASTEP) code. [38]This is crucial for the calculation of heavy elements, as reported in numerous theoretical studies. [39]The bandgap of γ-CsPbI 3 decreased with the Pb-I-Pb bond angle of 156.29°.The band edges were predominantly formed by the Pb and I states, indicating that the [PbI 6 ] 4− octahedral tilt plays a significant role in the formation of the energy states.This result demonstrates that the [PbI 6 ] 4− octahedral tilt can vary as a function of the position of the Cs atom in the 3D framework of the perovskite structure, resulting in minute changes in the unit cell volume.On the basis of our observations and results, we propose that changes in the optical bandgap of γ-CsPbI 3 are more likely to be caused by the insertion of the Cs ion into the 3D framework than by DMAI species remaining in the perovskite film after annealing.Owing to their hygroscopicity, [35] Cs ions can readily intercalate into the perovskite structure in the presence of moisture, enabling the formation of a distortion-relieved structure with a relatively narrower bandgap.We acknowledge that the calculation of band dispersion is valid, but the calculated bandgap is quantitatively underestimated because of the well-known PBE functional limitation.However, standard DFT can accurately predict atomic positions and structural energy, and evaluating the optical properties based on the optimized structures yielded similar insights and identical conclusions regarding a hybrid functional.Consequently, our computationally efficient calculations are suitable for determining the changes in the optical bandgap of γ-CsPbI 3 .
X-ray photoelectron spectroscopy (XPS) measurements were conducted to investigate the variation in binding energy with the atomic composition and interaction in perovskite films (Figure 3a,b).As a control experiment, we included CsPbI 3 perovskite film manufactured without DMA.No signal was detected for the N 1s peak at 402.2 eV in any of the films, indicating that DMA species are unlikely to have remained on the surface of the γ-CsPbI 3 perovskite films.The observed C 1s peak at 284.8 eV, corresponding to the C-C bond, was attributed to carbon adsorption and unintentional contamination during sample preparation.To confirm whether a change in chemical interaction occur in the samples, we determined the inorganic atomic components that were present.The Pb 4f 7/2 energy level of the control CsPbI 3 was 138.3 eV.In the case of the two films treated with DMA, the peak binding energy of Pb 4f 7/2 was shifted to 138.1 eV.This trend indicates that the binding energy of Pb-I interacts more strongly with γ-CsPbI 3 than with the control. [12,40]Interestingly, the binding energy associated with the formation of the perovskite lattice exhibited a different peak shift trend.The binding energies of the peaks of Cs 3d and I 3d for the control and inert condition-driven γ-CsPbI 3 were 619.1 eV for I 3d 5/2 and 724.7 eV for Cs 3d 5/2 .The Cs 3d and I 3d binding energies for the dry-driven γ-CsPbI 3 were decreased by 0.2 eV, confirming the stronger intercalation of Cs ions into the perovskite framework.The XPS analysis showed that the Cs cations could be intercalated into the perovskite structure under moist conditions, resulting in a widening of the Pb-I-Pb angle.
As shown in Figure 3c, we collected 1 Hnuclear magnetic resonance (NMR) spectra of the corresponding films to confirm whether there were changes in the amount of DMA residues as a function of the annealing conditions.We prepared the samples for 1 H-NMR analysis by dissolving the films directly in the DMSO-d 6 reference solvent in equal volumes and quantities for each subject (Figure S3, Supporting Information).This method allowed for improved sample preparation without contamination or powder loss.Although XPS did not reveal the presence of DMA residues, the 1 H-NMR spectra of the DMA-driven films exhibited distinct signals at δ = 8.15 and 2.55 ppm, which correspond to the -NH þ 2and -CH 3 protons, respectively. [13,16]The relative ratio of their detected spectra was 3 to 1.To quantify the amount of DMA residues, the corresponding regions were compared with the reference peak (Table S1, Supporting Information).Similar amounts of -NH þ 2and -CH 3 were found in the DMA-driven films, indicating that the humidity level during annealing had no effect on the amount of DMA residues in the perovskite films.Notably, we annealed the perovskite films at 220 °C, and the NMR results indicate that the remaining DMA residues are not the primary factor influencing the optical changes if the perovskite films are annealed at a temperature high enough to sublimate the DMA content over time.
All these findings demonstrate that the quality of γ-CsPbI 3 is strongly affected by moisture, which can control crystallization by manipulating the position of the Cs ions within the crystal lattice.To advance the fabrication systems, we further investigated the effect of the TEQ temperature on the γ-CsPbI 3 perovskite films under ambient conditions. [41]The TEQ conditions for annealing were determined by mapping the results of the Magnus formula and experimental data (Figure S4, Supporting Information).Figure 4a displays the J-V curves of the PSCs based on the γ-CsPbI 3 perovskite films annealed at various TEQ temperatures at 10% RH, as well as their device parameters.At TEQ temperature of −10 °C, the devices demonstrated the best PCE of 17.75%, which is primarily because of an improvement in the FF.This improvement may be attributed to the smaller parasitic resistive losses during charge carrier transport.SEM images were acquired to examine the morphologies of the perovskite films at each TEQ temperature to evaluate the relationship between the photovoltaic performance and morphology (Figure 4b).All films exhibited compact and pore-less film morphology.The grain size is compared in Figure 4c.The distribution of the crystal domain size was determined using a Gaussian model, with the peak maximum located at 669 nm at TEQ temperature of −5 °C, 920 nm at −10 °C, and 807 nm at −15 °C.As shown in Figure 4d, we compared the charge carrier lifetimes of γ-CsPbI 3 perovskite films using a time-correlated single-photon counting (TCSPC) system (Figure 4d). [42]Although the longest charge carrier lifetime of 11.99 ns was obtained at −10 °C of TEQ temperature, compared with the other temperatures, the lifetimes did not significantly improve under TEQ conditions, indicating that the bulk phase quality did not differ Energy Environ.Mater.2024, 7, e12625 strikingly.In addition, we performed PL lifetime image mapping to gain a deeper understanding of the γ-CsPbI 3 perovskite films (Figure 4e).We observed a clear distinction in the homogeneity of PL even though the brightness of the observed PL images did not change appreciably.In the −5 and −15 °C TEQ films, regions of inhomogeneity and dark spots were discernible in the PL images.Conversely, uniform PL and fewer dark spot areas appeared across the film at −10 °C TEQ, accounting for better charge carrier efficiency and the enhancement of the FF in the device.Accordingly, we confirmed that finetuning the RH level during the crystallization of γ-CsPbI 3 promoted uniform grain growth, which is closely related to the diffusion of Cs ions during ion exchange with DMA.
Although photovoltaic cells based on the optimized γ-CsPbI 3 film exhibited high efficiencies, as reflected by the J sc and FF parameters, the PCE was limited by a potential loss of ~0.62 eV in V oc relative to the maximum theoretical value.Considering the efficient parameters of J sc and FF, we anticipated that the inevitable incomplete band alignment caused by the material energy levels of TiO 2 ETL and γ-CsPbI 3 would lower the built-in potential (V bi ), the gap between the CBM of the ETL and the highest occupied molecular orbital of the HTL (i.e., CBM ETL − HOMO HTL ).As a result, we developed a-BaTiO 3 ETL, which is one of the perovskite oxides [43] as an alternative to the TiO 2 ETL to shift the CBM at the electron transport interface upwards.Figure 5a shows a schematic representation of the ETL fabrication process.Details are provided in the experimental section.Additionally, the XRD diffraction patterns are provided in Figure S5, Supporting Information.The XRD pattern of TiO 2 ETL exhibited that of anatase, while the ETL treated with Ba(OH) 2 solution had amorphous characteristics.The disorder of the lattice may be induced by Ba 2+ insertion and substitution.The chemical compositions of the deposited films were determined utilizing XPS. Figure 5b depicts the narrow XPS scan of the O 1s peak.Deconvolution of the O 1s peak revealed the various oxygen states present in the samples.Moreover, peaks associated with the lattice (O 2− ) were slightly shifted to a higher position by 0.01 eV (529.84 eV), and their area percentage in the O 1s spectrum appeared to be increased in a-BaTiO 3 .Thus, the non-lattice oxygen (O 1− ) state was suppressed in the a-BaTiO 3 , indicating TiO 2 has a more incomplete oxidation state.This result indicates that defects resulting from oxygen vacancies become less prevalent in a-BaTiO 3 .As shown in Figure 5c,d, the oxidation states of Ti are compared to further interpret the electronic state of the engaging elements.The Ba 3d peak was resolved into two components (Ba 3d 5/2 and Ba 3d 3/2 ), and the inset depicts the Ba peak of BaTiO 3 corresponding to Ba 3d 5/ 2 .The Ti 4+ in BaTiO 3 exhibited binding energies of 458.41 and 464.13 eV for the Ti 2p 3/2 and Ti 2p 1/2 peaks, respectively.Those of TiO 2 were observed at a higher energy level, 458.59 eV for 2p 3/2 and 464.31 eV for 2p 1/2 .This change was attributed to the transition of Ti 3+ to Ti 4+ to balance the oxygen state in the BaTiO 3 system.As a result, deconvolution of the Ti spectrum revealed that reduced Ti states were markedly suppressed in BaTiO 3 ; in terms of percentage, these Ti states constituted 19.7% of TiO 2 and 10.41% of BaTiO 3 .
Cs-corrected high-resolution scanning transmission electron microscopy (Cs-STEM) and electron energy loss spectroscopy (EELS) analyses were performed to investigate the atomic crystal structure and phase of a-BaTiO 3 (Target) in greater detail.Figure 6a depicts a Cs-STEM cross-sectional image of TiO 2 (Control).Despite the presence of a predominantly semicrystalline phase in the control, a defined crystalline periodicity with a lattice spacing of 3.51 Å was observed, which was indexed to the (101) plane of anatase TiO 2 .The (101) interplanar spacing derived from Cs-STEM images is in good agreement with the d-spacing values derived from the XRD results.In the Cs-STEM images of the target (Figure 6b), the absence of grain boundaries and periodic patterns indicated that Ba was successfully incorporated to form an amorphous BaTiO 3 phase.As seen in the inset, the corresponding fast Fourier transform (FFT) pattern is a halo ring, which represents the amorphous characteristics.Furthermore, the EELS results were compared to gain a deeper understanding of the chemical bonds between specific atoms and geometric sites.The near-edge structures in the L 3,2 edges of the EELS profile of Ti oxides correspond to covalent bond states resulting from direct and/or indirect interactions between O and Ti atoms. [44]igure 6c depicts the control L 3,2 energy loss near-edge structure (ELNES).In the case of a fully crystalline phase of TiO 2 , the L 3,2 edges exhibit a distinct four-peak pattern, and a lower oxidation state shifts the L 3,2 edges toward the lower energy loss position. [44]In this regard, the control appeared to have low crystallinity, and its oxidation states were highly dependent on the vertical position of the layer from the top, probably because of oxygen penetration during the annealing treatment. [45]Notably, the intensity is greater near the bottom than near the surface, owing to a dependence on the number of atoms to emit a signal.In addition, a second pre-shoulder appeared in front of the initial L 3 peak near the bottom, indicating the presence of species associated with a reduced phase of Ti x O y . [44,46]As shown in Figure 6d, the target demonstrated that the fingerprint of Ti 4+ became apparent with an additional peak to the left of the L 3,2 edges.The Ti-L 3,2 ELNES of the target displayed similar patterns without an additional pre-shoulder at the beginning of the first L 3 peak, regardless of the vertical position in the layer.Comparison of the Ti-L 3,2 ELNES reveals that the target has a single phase with fewer reduced Ti states, which results in a deep trap state.EELS mapping was also used to confirm the presence of engaging Energy Environ.Mater.2024, 7, e12625 elements.Figure 6e depicts the EELS mapping of the elements Ba, Ti, and O.The absence of any localized Ba signal confirms that Ba is alloyed with Ti and O. Consequently, the results confirmed that the developed a-BaTiO 3 as an alternative ETL for TiO 2 consisted of a single phase as opposed to Ba-doped TiO 2 hybrid nanoclusters [47] or separate domains of TiO 2 and BaTiO 3 phases. [48]he γ-CsPbI 3 absorber was deposited on each ETL for the fabrication of PSCs. Figure 7a displays the SEM cross-sectional images of the respective PSCs.The measured thickness of the perovskite layers was ~700 nm, which means that the layers were dense in both cases.
We characterized the J-V measurements under the illumination of one sun.The best-performing devices are compared in Figure 7b.The device with the TiO 2 ETL displayed a PCE of 17.98% with photovoltaic parameters of a V oc of 1.097 V, J sc of 20.53 mA cm −2 , and FF of 79.8%.The detailed forward and reverse J-V parameters are shown in Table S2, Supporting Information.The device performance was primarily constrained by a potential loss in V oc .The device with the a-BaTiO 3 ETL produced a remarkable PCE of 19.96%, which was primarily achieved by an increase in V oc , yielding a V oc of 1.167 V, J sc of 20.91 mA cm −2 , and FF of 81.8%. Figure 7c depicts the steady-state efficiencies of the devices as measured at the maximum power point voltage under the illumination of one sun.Tens of seconds after turning on the illumination, the maximum power efficiency of the control device reached 17.1%.By contrast, the target device was capable of instantaneously achieving the maximum power efficiency of 19.6%. Figure 7d depicts the external quantum efficiency (EQE) outcomes of the respective devices.The integrated J sc from the EQE spectra for the control and the target was 20.40 and 20.73 mA cm −2 , respectively.Higher conversion efficiencies in the target were measured at wavelengths lower than 500 nm, attributable to an increase in carrier lifetimes, and a signature of improved carrier extraction was observed at the edge of the EQE. Figure 7e,f depict the statistical distributions of the devices based on specific photovoltaic parameters.The average and standard deviation values are provided in Table S3, Supporting Information.Their efficiency distribution was narrow, presumably because of the well-deposited perovskite layer, and the photovoltaic parameters showed a marked improvement in V oc .
As depicted in Figure 8a, we performed V oc rise (OCVR) and V oc decay (OCVD) measurements to further elucidate the relationship between the quick transient response in the current and enhanced V oc . [49]Upon illumination, quasi-Fermi level splitting takes place, and photovoltage is generated through the redistribution of the internal electric field.The transient times to reach the maximum point depend on the equilibrium state of the internal electric field, whose latency is highly sensitive to shallow trap states. [50]Although the maximum voltage (V max ) of the target device was reached within a few seconds, the voltage of the control device gradually increased over time to reach V max , indicating that the light soaking had a trap-filling effect. [50,51]ccording to the results of the OCVD measurements, the photogenerated carriers in the target require more time to disappear through recombination.To account for the enhancement of the target device, a series of characterizations were conducted in which J-V was measured at varying light intensities (Figure 8b).The photovoltaic plot had a sloped scale.In comparison to that of the control device (1.385 kT/q), the slope of the target device was 1.172 kT/q, where k is the Boltzmann constant, T is the absolute temperature, and q is the base charge.A slope close to the unit value of 1 indicates the absence of nonradiative recombination in the device owing to the deep trap states, confirming that the target device has suppressed nonradiative recombination. [52]In addition, the linearity of J sc as a function of light intensity resulted in a slope of 0.9969 for the target device, which is approximately one unit, indicating superior charge carrier extraction from the carrier transport interface relative to the control.Electrochemical impedance spectroscopy (EIS) was performed by applying bias and light stimuli to simulate quasi-steady-state conditions to quantitatively compare their carrier lifetimes. [50]Figure 8c provides Nyquist and Bode plots with the equivalent circuit shown in the inset.The high-frequency impedance spectroscopy (IS) response near V oc provides information regarding the nonradiative recombination of PSCs, whereas the IS response at low frequencies is correlated with the dynamic response of mobile ions. [50,53]In the high-frequency range, a significant difference in the recombination resistance (R rec ) was recorded, and their component was evaluated using τ = 1/(2πf m ), where f m is the frequency at the peak maximum of the Bode plot.The extracted carrier lifetimes of the target and control devices were 1004.2 and 710.0 ns, respectively.The EIS measurements confirmed that a-BaTiO 3 had a charge carrier transport advantage close to that of the inorganic perovskite layer.
To comprehend the enhancement of V oc in devices with the a-BaTiO 3 ETL, we characterized the V bi using Mott-Schottky (MS) analysis.To obtain a dependable depletion layer capacitance (C dl ), the devices were fully relaxed in the short-circuit state in the dark. [26]As depicted in Figure 9a, the capacitance was determined as a function of the bias voltage (V app ), which comprised three capacitive features.The linear fit of C dl is typical of the prevalent relationship C À2 dl ¼ 2 V bi ÀV ð Þ=qεε 0 N, where q, ε, ε 0 , and N are the elementary charge, relative dielectric constant, vacuum permittivity, and defect density, respectively.V bi can be extracted from C dl using the linear fit employed to analyze MS plots. [54]he device containing a-BaTiO 3 ETL exhibited a V bi of 1.23 V, which was 70 mV higher than that of the control.The increase in V bi indicates that the a-BaTiO 3 ETL can modify the energy band alignment for electron transport to generate a higher V oc .Figure 9b depicts the ultraviolet Energy Environ.Mater.2024, 7, e12625 photoelectron spectroscopy (UPS) results of the TiO 2 ETL and a-BaTiO 3 ETL to determine the relationship between the observed trends in V bi and the energy band profiles.The secondary cutoff spectra indicated that the work function (WF) of a-BaTiO 3 was altered, corroborating the contribution from interactions between the Ti d and O p orbitals as a result of Ba insertion, which is in good agreement with the XPS and EELS results.Figure 9c depicts the band profiles of the ETLs.The CBM was determined by subtracting the bandgap established by the Tauc plots of UV-vis spectra in Figure S7, Supporting Information.The bandgaps derived from the UV-vis spectra corresponded well with the reported values.The band profile demonstrates that a-BaTiO 3 was able to successfully modify the energy band alignment of the device.To obtain the landscape of energy band alignment in the devices, we conducted UPS analysis on the perovskite layer (Figure S8, Supporting Information).Figure 9d illustrates a schematic diagram of the band alignment in the devices based on TiO 2 and a-BaTiO 3 as ETLs.Although the cascaded band alignment in these devices, the driving force behind the band offset (ΔG) for the charge extraction potential is highly dependent on the ETL, where ΔG = Perovskite CBM − ETL CBM .The potential loss of the TiO 2 and a-BaTiO 3 ETLs was calculated to be 0.42 and 0.19 eV for charge extraction, respectively, which indicates efficient electron extraction in the device with a-BaTiO 3 ETL.Accordingly, the improved performance can be attributed to the low potential loss of charge extraction in devices based on a-BaTiO 3 .We further evaluated the systematic advantages of using a-BaTiO 3 as the ETL in inorganic PSCs by comparing our results with ΔG values from other reports.Figure 9e depicts a systematic comparison of ΔG utilizing TiO 2 as ETL in inorganic PSCs, demonstrating the superiority of a-BaTiO 3 over TiO 2 for application to inorganic PSCs as ETLs.We observed that devices containing a-BaTiO 3 showed excellent stability under ambient conditions and retained 95% of their initial efficiency after 2000 h of exposure to the air without encapsulation.Moreover, the devices with a-BaTiO 3 exhibited a slower PCE decay under full-spectrum illumination after 100 h in the air without encapsulation than those with TiO 2 (Figure 9f).The initial J-V parameters are provided in Table S4, Supporting Information.

Conclusions
This study demonstrated a straightforward and effective method for realizing the full potential of CsPbI 3 .The bandgap and crystal strain of γ-CsPbI 3 perovskite were altered by the diffusion of Cs ions for crystal reconstruction via ion exchange, which affected the bulk defects.The optoelectronic quality of γ-CsPbI 3 was highly dependent on the presence of moisture during γ-CsPbI 3 film formation, and we determined the optimal environmental conditions for obtaining high-quality γ-CsPbI 3 perovskite films with micron-sized grains and low trap density.Additionally, the alternative ETL, amorphous BaTiO 3 exhibited a   based PSCs demonstrated greater stability under full-sunlight illumination than TiO 2 -based PSCs.Thus, the incorporation of the a-BaTiO 3 ETL is expected to play a significant role in maximizing the potential of inorganic PSCs.
Fabrication of electron transport layer: Fluorine-doped tin oxide (FTO) glass was washed for 5 min under sonication with acetone, ethanol, and 2-propanol (IPA).The TiO 2 layer was deposited via chemical bath deposition (CBD) of a TiCl 4 aqueous solution.The CBD of TiCl 4 aqueous solution was carried out by immersing oxygen plasma-treated FTO glass in a 200 mM TiCl 4 solution at 70 °C for 45 min, then rinsing the FTO glasses with IPA and drying them in a stream of N 2 .The TiO 2 -coated substrates were thermally treated at 100 °C for 30 min, followed by 520 °C in the air for 30 min.For a-BaTiO 3 deposition, a stock solution of Ba (OH) 2 was prepared by dissolving 4 mg mL −1 of Ba(OH) 2 powder in deionized water.The a-BaTiO 3 layer was formed by spin-coating the Ba(OH) 2 solution onto the as-deposited TiO 2 film at 3000 rpm for 30 s and then thermally treating the film for 30 min at 520 °C in air.
Fabrication of solar cells: The FTO glass coated with metal oxide was exposed to UV/O 3 for 5 min before being transferred to a glove box filled with nitrogen.A perovskite precursor solution containing 1.0 M PbI 2 , 1.3 M DMAI, and 1.0 M CsI was prepared by dissolving these powders in a solvent mixture of DMF and dimethyl sulfoxide (DMSO) (volume ratio = 4:1).The perovskite absorber layer was deposited using a sequential program set at 3000 and 6500 rpm for 30 and 10 s, respectively.The freshly spun film was transferred with care to an air glovebox equipped with RH and temperature controls and then annealed at 220 °C for 5 min.To deposit the passivation layer, an HTABr solution (1,2dichlorobenzne:IPA = 97:3 (vol%)) was spin-cast at 5000 rpm for 30 s on top of the perovskite layer.The resultant film was thermally treated for 10 min at 100 °C.The spiro-OMeTAD (72.3 mg mL −1 in chlorobenzene) solution mixed with 28.8 μL tBP, 17.5 μL Li-TFSI (520 mg mL −1 in acetonitrile), and 29 μL FK209 (300 mg mL −1 in acetonitrile) was then spin-cast at 3000 rpm for 30 s onto the perovskite film.Finally, a 70 nm layer of Au electrode was thermally evaporated.
Characterization: The current density-voltage (J-V) curves of PSCs were recorded using a 0.096 cm 2 metal mask and an electrochemical station (VSP; Bio-Logic) under one-sun illumination (100 mW cm −2 AM1.5G) from a solar simulator (Sun 3000, Class AAA; ABET Technology) under the atmospheric temperature of 25 °C and RH of 25 AE 10% condition.The illumination was calibrated using a KG5-filtered Si reference cell (Newport).The scan rate was 40 mV s −1 , and the voltage step was 10 mV.The stability test was conducted under the Energy Environ.Mater.2024, 7, e12625 same illumination condition.The atmospheric temperature was 38 AE 2 °C, and the RH was 25 AE 10%.No ultraviolet filter and cell encapsulation were applied.Additionally, the EQE spectra were obtained using a potentiostat (Ivium) and a monochromator (Dongwoo Optron Co., Ltd) under a Xe lamp (ABET 150 W, AET) in a dark box.The high-resolution XRD patterns were acquired at a scan rate of 2°min −1 (SmartLab 9 kW; Rigaku).In addition, the morphologies of samples were investigated by FE-SEM (SU8010; Hitachi).UPS was performed using a He I instrument (21.2 eV) (Thetaprobe; Thermo Fisher), and the energy scale was calibrated to the Fermi-edge of a sputtered Au sample.A monochrome X-ray source of Al-Kα with a full width at half maximum of 0.1 eV was employed to assist XPS measurements, with 200 runs for the survey scan and 50 runs for the narrow scans.The V oc curves were obtained using the BioLogic instrument in express mode.Additionally, IS responses were obtained using the Ivium potentiostat under the illumination of 50 mW at an applied voltage of 0.9 V and fitted using the ZView software.A diode-pumped solid-state laser (Omicron) with an excitation wavelength of 532 nm was used to acquire the steady-state PL spectra.In addition, the absorption spectra were recorded using a UV-vis-NIR spectrophotometer (Cary 5000; Agilent Technologies).The 1 H-NMR analysis was carried out by using a Bruker Avance III 600 MHz system equipped with a 5 mm TCI ( 1 H/ 13 C/ 15 N) cryoprobe and DMSO-d 6 as the reference solvent.To prepare the 1 H-NMR samples, we dissolved 9 of each target condition in DMSO-d 6 solvent.
Transmission electron microscopy analysis: Cs-STEM images were obtained using a JEM-ARM200F electron microscope operated at 200 kV.The crosssectional-STEM samples were prepared using the focused ion beam (FIB) method with a Helios G5 UC, and amorphous carbon paste was coated on the ETL for EELS analyses with a Model 965 GIF Quantum ER.Cs-STEM analyses were conducted on JEOL Ltd JEM-ARM200F instrument installed at the National Center for Inter-university Research Facilities (NCIRF) at Seoul National University.
Time-resolved photoluminescence decay and imaging measurement: The time-resolved photoluminescence study was carried out using a confocal microscope (MicroTime-200; Picoquant, Germany) with a 40× objective.The lifetime measurements were performed at the Korea Basic Science Institute (KBSI), Daegu Center, Korea.A single-mode pulsed diode laser (470 nm with a pulse width of ~30 ps and an average power of ~20 nW) was used as the excitation source.Using a dichroic mirror (490 DCXR, AHF), 75 μm pinhole, long-pass filter (FEL0700; Thorlabs), and single-photon avalanche diode (PDM series, MPD), emission from the samples was collected.The emission photons were counted using a TCSPC system (PicoHarp300; PicoQuant GmbH, Germany).The PL lifetime images consisted of 200 × 200 pixels, which were recorded using the time-tagged timeresolved (TTTR) data acquisition method.Exponential fitting for the obtained emission decays was accomplished using the Symphotime-64 software (Ver.2.2).
Computational methods: The first-principles DFT calculations were performed using the CASTEP plane-wave pseudopotential code.The cores were handled with on-the-fly generated (OTFG) ultrasoft pseudopotentials.The GGA-PBE exchange-correlation functional was used by employing noncollinear and SOC.Moreover, the geometry was optimized using the Broyden-Fletcher-Goldfarb-Shannon (BFGS) algorithms.The parameters evaluated for structural optimization and energy were set to converge in terms of energy, force, and displacement at 1.0 × 10 −5 eV atom −1 , 0.003 eV Å−1 , and 0.001 Å, respectively.Both the global orbital cutoff and the self-consistent field (SCF) were set to 570 eV and 1.0 × 10 −6 eV atom −1 , respectively.Fitted with a split pseudo-Voigt function, the orthorhombic γ-CsPbI 3 phase from the diffraction patterns was used to develop a structural model.Additionally, the Visualization for Electronic and Structural Analysis (VESTA) was utilized for 3D visualization.

Figure 1 .
Figure 1.a) XRD patterns of the perovskite films prepared from different atmospheric conditions.Crystallography calculations of b) unit cell volume and c) crystal strains of William-Hall plot.d-h) SEM images of the corresponding perovskite films.i) PL spectra of the corresponding perovskite films deposited on FTO.

Figure 3 .
Figure 3. XPS spectra of a) full survey scan and b) narrow scan on γ-CsPbI 3 .c) 1 H-NMR spectra of samples obtained from the corresponding films by directly dissolving in DMSO-d 6 ."Inert" and "w/o DMA" are representative of Nitrogen and <1% of RH conditions, and 10-15% of RH conditions for "Dry".

Figure 4 .
Figure 4. a) J-V curves of the PSCs based on perovskite films prepared at different TEQ temperatures.The masked active area is 0.096 cm 2 .b) SEM images, c) grain size distribution, d) TCSPC measurement results and calculated carrier lifetime values, and e) PL images of the corresponding perovskite films.

Figure 6 .
Figure 6.Cs-STEM images of a) TiO 2 and b) a-BaTiO 3 ; the insets show the corresponding selected area FFT pattern.EELS spectra for Ti L 2,3 edges of c) TiO 2 and d) a-BaTiO 3 ; the background was subtracted.e) Cs-STEM image and EELS mapping images for Ba, Ti, and O in a-BaTiO 3 .

Figure 7 .
Figure 7. a) Cross-sectional SEM images of PSCs based on TiO 2 (control) and a-BaTiO 3 (target) as the ETL.b) J-V curves of the best-performing PSCs and c) Maximum power point current and efficiency at 0.91 (control) and 0.97 V (a-BaTiO 3 ).The masked active area is 0.096 cm 2 .d) EQE spectra of the corresponding devices.Distribution of the fabricated PSCs with respect to the photovoltaic parameters: e) efficiency and FF, and f) V oc and J sc .

Figure 8 .
Figure 8. a) OCVR and OCVD measurement.b) Light intensity dependence of the PSCs on photovoltage and photocurrent.c) Nyquist and Bode plots recorded at a bias of 1 V under illumination of 50 mW m −2 ; the inset shows the equivalent circuit used to fit the plots.

Figure 9 .
Figure 9. a) Capacitance-voltage (C-V) curves.b) UPS profiles of TiO 2 (control) and a-BaTiO 3 (target) ETL.c) Schematic energy level diagram of TiO 2 and a-BaTiO 3 films with respect to the E vac vacuum level (IE, ionization energy; E gap , bandgap).d) Band profiles of PSCs based on TiO 2 and a-BaTiO 3 .e) Comparison of the band offset in inorganic PSCs.The circle and star symbols represent TiO 2 and a-BaTiO 3 ETL-based devices, respectively.f) Long-term stability results of the corresponding devices without encapsulation (RH 25 AE 10%).