Remarkable progress has been made in perovskite solar cells (PSCs) since the first report of all-solid-state PSCs in 2012^[1], with certified power conversion efficiencies (PCEs) reaching 28.0% in 2026^[2]. Formamidinium lead iodide (FAPbI₃) perovskites have emerged as leading photoactive compositions^[3], owing to their near-ideal bandgap^[4], strong light harvesting^[5], high defect tolerance dominated by shallow trap states^[6], and effective dielectric screening of charge carriers arising from the soft lattice, which ensures long carrier diffusion lengths and lifetimes^[7]. Despite these favorable optoelectronic properties, the structural instability of the perovskite phase remains a persistent challenge for halide perovskite materials^[6]. FAPbI₃ is intrinsically susceptible to moisture^[8], undergoing several decomposition pathways: the α-to-δ phase transition (to a yellow, photoinactive phase)^[9], decomposition into PbI₂ and formamidinium iodide (FAI)^[10], loss of the organic FA⁺ cation^[11], and the formation of iodine-related defects^[12]. Moreover, the interplay between perovskite defects near the buried interface and surface defects of the underlying metal oxide (e.g., SnO₂) has been shown to initiate perovskite degradation^[13,14], leading to severe performance loss and compromised long-term stability. SnO₂ is among the most widely used electron transport layers in n-i-p-structured PSCs, owing to its high transparency, excellent electron mobility, low-temperature processability, and favorable band alignment with the perovskite absorber^[15]. However, the SnO₂ surface contains a variety of defects and adsorbates - including oxygen vacancies, hydroxyl groups, dangling bonds, and chemisorbed oxygen - which can accelerate both nonradiative recombination and moisture-induced interfacial degradation^[14]. Vacancy-related defects are common in solution-processed SnO₂ and create reactive interfacial sites for water molecules under environmental stress^[14,16,17]. Oxygen vacancies in SnO_2-x, promoted by adsorbed H₂O, raise the positive charge density on neighboring Sn atoms and thereby elongate and weaken the interfacial Pb-I bond. The weakened bonding accelerates interfacial ion migration and ultimately drives phase decomposition^[12,18,19]. This destabilized bonding environment lowers the barrier for iodine interstitial formation and promotes the diffusion of iodine-related defects into the defect-rich SnO₂. Furthermore, oxygen-deficient SnO₂ thermally activates oxygen migration^[20], and an applied electric bias further drives the release and redistribution of oxygen-related species from SnO₂ toward the perovskite, aggravating interfacial instability^[21]. Oxygen vacancies can additionally serve as bonding sites for A-site cations in halide perovskites (APbX₃; A⁺ = monovalent cation and X^- = halide)^[22]. The resulting structural distortion, combined with FAI deprotonation and iodine-defect generation, favors transitions from the photoactive α-phase to non-photoactive δ-FAPbI₃, PbI₂, and other secondary phases^[22,23]. Interface engineering of the SnO₂/FAPbI₃ buried interface is therefore critical to phase stability of perovskite^[16], suppressing the defect-mediated degradation pathways. In this regard, various interfacial engineering strategies have been reported to suppress defects at the buried interface^[24-26]. In particular, alkali halides have been widely employed as interlayers at metal oxide/perovskite interfaces^[24,27,28]. One of the most common interlayers based on alkali halides is KCl, which effectively passivates interfacial defects to suppress charge recombination and simultaneously regulates crystallization of perovskite films^[27]. Similarly, CsCl has been employed either as an interlayer at the buried interface or as an additive in the perovskite precursor^[29-31]. CsCl as an interlayer was found to improve the resistance of methylammonium lead iodide (MAPbI₃) to ultraviolet light and to enhance the moisture stability of FAPbI₃ by reducing the microstrain of the overlying perovskite film^[32,33]. In addition, CsCl has been widely employed as an additive in the perovskite precursor solution, where the partial substitution of FA⁺ by smaller Cs⁺ cations introduces beneficial lattice strain and thereby stabilizes α-phase^[33]. The phase-stabilizing role of CsCl in FAPbI₃ has been attributed to Cs⁺-induced contraction of the octahedral cages and modification of the Pb-I bonding environment, which relieves the excess lattice strain of pristine FAPbI₃^[22]. The accompanying strengthening of the Pb-I framework raises the thermodynamic stability of the α-phase and lowers the δ-to-α transition temperature^[18,20]. The improved moisture resistance reported for FAPbI₃ films deposited on a CsCl interlayer has likewise been rationalized within this framework^[21,33]. CsCl has thus been reported to be effective both as an interlayer and as a precursor additive, although its role has largely been limited to lattice stabilization and/or improved crystallinity of perovskite films.

In this study, we focus instead on the structural resilience of FAPbI₃ grown on a SnO₂ layer, examining how the α-phase recovers after moisture-induced decomposition in the presence of a CsCl interlayer and separating this interfacial effect from phase stabilization within the bulk FAPbI₃ film. Structural resilience, defined here as the recovery of the α-phase lattice following stress-induced degradation, is therefore distinct from the phase stability or the degradation itself that previous reports have mainly addressed. Here we introduced a CsCl interlayer at the SnO₂/FAPbI₃ buried interface of n-i-p-structured PSCs to mitigate the coupling effect of moisture- and thermal stress-driven phase degradation originating at this buried interface. We systematically examined how the CsCl interlayer reinforces the humidity resilience of α-phase by stabilizing the vulnerable SnO₂/FAPbI₃ buried interface, taking into account the coupled contributions of the underlying SnO₂ substrate and the overlying FAPbI₃ to phase stability against moisture. Beyond the well-known reversible α-to-δ transition, FAPbI₃ exposed to 82%-88% relative humidity (RH) underwent irreversible decomposition to Pb-O-related species upon post-exposure annealing. The CsCl interlayer suppressed formation of both oxygen- and iodine-related defects at the buried interface, allowing the post-exposure annealing to partially recover the α-phase from the δ-phase and PbI₂ with retained FA⁺.

A CsCl interlayer was employed at the SnO₂/FAPbI₃ buried interface in n-i-p (SnO₂/FAPbI₃/spiro-OMeTAD)-structured PSCs, with the CsCl solution concentration varying from 10-90 mM [Figure 1 and Supplementary Figure 1]. Devices without the CsCl interlayer, hereafter referred to as the control, were compared with the CsCl-treated counterparts. The short-circuit photocurrent density (J_SC) was largely independent of CsCl treatment [Figure 1B and Supplementary Figure 1B], indicating that the CsCl interlayer has little effect on the light-harvesting efficiency of the perovskite layer. Peak performance was reached at 50 mM, which is taken as the target condition hereafter. The target device delivered the highest open-circuit voltage (V_OC) of 1.105 V, against 0.918 V for the control device [Figure 1C and Supplementary Figure 1C], and the best PCE increase from 17.15% for the control device to 22.15% for the target device [Figure 1E and Supplementary Figure 1E]. The stabilized power outputs of the control and target devices [Supplementary Figure 2] support the PCE gain from the CsCl interlayer. The higher V_OC and maximum-power-point voltage (V_mpp) of the target devices reflected suppressed nonradiative recombination at the interface, as evidenced by the largest recombination resistance (R_rec) over the CsCl concentration range of 10-90 mM [Supplementary Figure 3]. The voltage-dependent R_rec is shown in Supplementary Figure 3A by fitting the Nyquist plots [Supplementary Figures 3B-E] to an equivalent circuit. The low-frequency semi-arc is commonly assigned to defect-mediated recombination at the perovskite interface^[34], indicating that CsCl passivates surface defects acting as recombination centers on both the SnO₂ and FAPbI₃ surfaces. Cl^- ions released from the CsCl interlayer likely coordinate with undercoordinated Pb²⁺ and halide vacancies on the FAPbI₃ side and with oxygen-related defects on the SnO₂ side^[35,36]. Both interactions lower the overall interfacial defect density and suppress the defect-mediated charge recombination at the buried interface. The beneficial effect of the CsCl passivation is also evident in the long-term shelf stability test, where the photovoltaic performance of the target devices was better maintained over 2,000 h of dark storage than that of the control devices [Supplementary Figure 4].

The effect of CsCl-treated SnO₂ substrates on the crystal growth of FAPbI₃ was then examined by XRD. XRD patterns of the control and target perovskite films are shown in Figure 2A. Both perovskite films grew preferentially along the (001)/(002) direction, with a higher (001) intensity observed for the target film. The stronger preferential orientation indicates improved charge extraction along the vertically aligned [PbI₆]^4- framework and fewer defects from reduced structural disorder^[37], consistent with higher R_rec of the target devices^[38]. The perovskite film crystallinity is generally governed by nucleation and growth kinetics, both of which depend on the underlying interfacial chemistry. Alkali halide interlayers such as KCl have been shown to slow crystallization through interactions between perovskite precursors and ions released from the interlayer^[39,40]. Likewise, Cl^- transiently coordinates with Pb²⁺ during crystallization, favoring oriented growth. This enhanced structural orientation is also reflected in film morphology. The grain size distributions of the control and target films are compared in Figure 2B, where the target film produced larger grains. Although the coverage of the extremely thin CsCl interlayer is difficult to confirm directly, the uniform morphology and full coverage of the overlying perovskite films [Supplementary Figure 5] suggest that the CsCl interlayer is uniformly distributed, although this evidence is indirect. Grain size increased with CsCl concentration up to 50 mM and then mostly saturated at 90 mM [Supplementary Figures 5 and 6]. The (001) diffraction peak shifted slightly from 14.20^o in the control film to 14.23^o in the target film. This shift toward higher angle reflects lattice contraction, consistent with the incorporation of Cs⁺ released from CsCl into the perovskite lattice during crystal growth^[33]. To probe the depth dependence of this contraction, GIXRD patterns were collected at tilt angles (Ψ) between 10^o and 40^o [Supplementary Figure 7], where larger Ψ reflects crystallographic information from greater depths beneath the top surface. The corresponding 2θ-sin²Ψ plot is shown in Figure 2C, where the target film shows a consistently higher 2θ than the control across the full Ψ range, reflecting the reduced lattice spacing that follows from substitution of FA⁺ by the smaller Cs^+[41,42]. This confirms that lattice contraction persists throughout the film depth and indicates that Cs⁺ diffusion toward the top surface was facilitated during crystal growth. The depth-independent contraction, on the other hand, had little effect on the residual lattice strain of the bulk perovskite, as reflected by comparable slope of the 2θ-sin²Ψ plot^[43]. Both films retain tensile strain of comparable magnitude, despite the overall lattice contraction in the target film.

To evaluate the effect of the CsCl interlayer on the moisture resistance of FAPbI₃ films, XRD analysis was performed at three stages: pristine film, the film after humidity exposure at 82-88% RH, and the film after post-exposure annealing at 170 °C intended to restore the α-phase from the humidity-induced δ-phase. The relative humidity was set to 82%-88% RH, close to the 85% RH specified in the ISOS-D-3 damp-heat protocol^[44,45]. Figure 3 presents the structural degradation and subsequent recovery of the (001)-oriented control and target FAPbI₃ films, alongside the corresponding optical images. As shown in Figure 3A, the control film exhibited a substantial decrease in the (001) diffraction intensity after humidity exposure, accompanied by the evolution of the δ-phase and pronounced PbI₂ peaks, consistent with the black-to-yellow color change of the film. Unlike MAPbI₃, whose degradation under humid conditions is often associated with hydrate phase formation^[46], the degradation of FAPbI₃ has more commonly been described in terms of the α-to-δ phase transition and decomposition into FAI and PbI₂^[10]. A quantitative comparison of the α-phase, δ-phase, and PbI₂ peak intensities across the three stages is shown in Supplementary Figure 8. Diffraction peaks corresponding to both δ-phase and PbI₂ emerged after humidity exposure [Supplementary Figure 8B and C], indicating that degradation proceeded through both pathways. Although the post-exposure annealing eliminated the humidity-induced δ-phase [Figure 3A and Supplementary Figure 8B], the α-phase barely recovered and the control films turned white, indicating irreversible degradation driven by the combined effect of moisture and thermal stress. The buried interface typically hosts a high density of defects - oxygen vacancies, hydroxyl groups, iodide vacancies, and undercoordinated Pb sites - that readily trap water molecules under humid conditions and accelerate interfacial degradation^[14,36,47], as evidenced in Supplementary Figure 9. The FAPbI₃ film deposited on the SnO₂-coated FTO substrate (control film) showed poor moisture resistance during humidity exposure and underwent further irreversible conversion during the post-exposure annealing [Supplementary Figure 9A], whereas both processes were largely suppressed in the film grown on the bare FTO substrate [Supplementary Figure 9B]. Figure 3B shows the corresponding stages for the target film. The initial (001) intensity is higher than that of the control, consistent with the role of CsCl in promoting preferential (001) orientation during growth^[29,48-50]. Despite the apparent reduction of the (001) peak during humidity exposure and the emergence of δ-phase and PbI₂ peaks, as also seen in the control film, the target film retained considerable (001) intensity, indicating that the CsCl interlayer improves the moisture tolerance of FAPbI₃ films. Remarkably, the target film partially restored the α-phase from the humidity-induced δ-phase during the post-exposure annealing. However, full recovery was not achieved despite complete removal of the δ-phase during the post-exposure anneal. In Supplementary Figure 10, several metal halides were incorporated as an interlayer and examined in the same manner to separate the individual contributions of Cs⁺ and Cl^-. The choice of interlayer altered the crystallinity of the overlying perovskite film, and the metal chlorides (NaCl, KCl, and CsCl) gave better crystallinity than the metal iodide (CsI). Although the α-phase recovery is less pronounced than in the CsCl-based target films, these interlayers still produce a partial recovery during the post-exposure annealing [Supplementary Figure 11], suggesting that metal halide interlayers reduce oxygen-vacancy-related defects at the SnO₂ interface, albeit to differing extents depending on the interlayer, and suppress moisture-induced interfacial degradation. This interlayer-driven effect is distinct from the benefit typically attributed to high bulk crystallinity^[51,52], as evidenced by the pronounced (001) peak loss in Supplementary Figure 12. When CsCl was instead introduced as a precursor additive, its effect on crystal growth was more pronounced, yielding a sharper (001) peak than that of the target film. After humidity exposure, however, the (001) intensity dropped sharply, confirming that the moisture stability of the target film originates primarily from interfacial effects rather than from bulk crystallinity, although the CsCl-induced improvement in bulk crystallinity still makes a smaller contribution. The structural resilience to moisture therefore appears to be maximized by synergistic effects of passivation of the SnO₂ surface through interlayer coordination and improved crystallinity of the overlying perovskite film, with the CsCl interlayer showing a more pronounced synergistic effect than the independent effects of Cs⁺ or Cl^-.

FAPbI₃ consists of a corner-sharing [PbI₆]^4- octahedral framework, with organic FA⁺ cations occupying the cuboctahedral cavities and coupling to the inorganic cage through a combination of ionic and noncovalent interactions^[7]. Under moisture exposure, water preferentially hydrogen-bonds to FA^+[53], destabilizing the N-H···I interaction and weakening the FA⁺-[PbI₆]^4- coupling. FT-IR spectroscopy was performed at the three sequential stages to track the molecular bonding environment within the perovskite lattice. For both the control and target pristine films [Figure 4], the characteristic FA⁺ vibrational features appear at 3,400-3,200 cm^-1, ~1,712 cm^-1, and ~1,354 cm^-1, corresponding to the N-H stretching, C=N stretching, and C-H vibrational modes, respectively^[54]. Quantitative analysis of the N-H stretching peak is shown in Supplementary Figure 13. For the control sample, the N-H stretching mode showed pronounced broadening and attenuated intensity to ~36.4% of its pristine value after humidity exposure [Figure 4B and Supplementary Figure 13A], indicating that the original N-H···I interactions within the FA⁺-[PbI₆]^4- sublattice were substantially disrupted, likely through the formation of strong H₂O···FA⁺ hydrogen bonds. The C=N-related vibrational peak, by contrast, remained largely intact after humidity exposure, confirming that the FA⁺ molecular backbone itself was preserved. Even after post-exposure annealing, the FA⁺-[PbI₆]^4- interaction was barely restored but rather further weakened, to ~25.5% of the pristine value, suggesting that the FA⁺ cations remained preferentially bound to H₂O rather than the inorganic cage. In contrast, the target sample retained more than 55.6% of its pristine N-H intensity throughout the humidity exposure and post-exposure annealing [Figure 4D and Supplementary Figure 13B]. The CsCl-induced stabilization of the buried interface, which would otherwise act as a preferential adsorption and trapping site for H₂O molecules^[12], suppressed moisture accumulation at this defect-rich region, thereby preserving a pristine-like N-H···I bonding environment within the FA⁺-[PbI₆]^4- framework. In addition to the interfacial effect, Cs⁺-induced α-phase stabilization may also contribute to maintaining the FA⁺-[PbI₆]^4- interaction to some extent^[22,52], as pointed out earlier.

Moisture-induced degradation of perovskite films typically proceeds via water adsorption at the film surface, diffusion through grain boundaries, and accumulation at the hydrophilic buried interface^[14,55], which is widely considered as the most defect-rich region of n-i-p-structured PSCs^[56]. Water desorption from this interface requires relatively high energy (~0.70 eV)^[12] due to strong interactions with the surrounding defect sites, indicating direct binding of H₂O to Pb- and I-related defects on the perovskite side and to O-related sites on the SnO₂ surface during moisture-induced phase degradation of FAPbI₃. XPS was therefore performed to track changes in the chemical state of the control and target films at three sequential stages. Supplementary Figure 14 shows the depth-resolved XPS profiles of the I 3d_5/2 peak. The control sample exhibited a clear depth dependence of the I 3d_5/2 binding energy after post-exposure annealing [Supplementary Figure 14A], whereas the target film showed only minimal binding-energy shifts across all three stages and across the full depth range [Supplementary Figure 14B]. The largest binding-energy change in the control sample appeared at a depth of 400 nm, near the buried interface, confirming that humidity-induced degradation during the post-exposure annealing was driven primarily by moisture accumulated at this interface. XPS spectra of the I 3d_5/2 and O 1s peaks near the buried interface are shown in Figure 5, with dashed guidelines indicating the pristine binding-energy positions. The I 3d_5/2 envelopes [Figure 5A and B] were deconvoluted into two components at each stage: I^- in FAPbI₃ lattice (red, 618.74 eV) and I^- in PbI₂ (blue, 619.32 eV)^[57,58]. In the pristine stage, the target film showed a higher FAPbI₃/PbI₂ ratio than the control, consistent with its higher crystallinity. The coexistence of PbI₂ with FAPbI₃ in both pristine films is also supported by a weak PbI₂ XRD peak [Figure 2A] and by the white PbI₂ flakes visible in the surface SEM images (Figure 2B, inset)^[59]. Upon humidity exposure, moisture accelerated the transition to the δ-phase (Eq. 1)^[10,12], while both samples produced additional PbI₂ via partial FAPbI₃ decomposition (Eq. 2), accompanied by a shift of the I 3d_5/2 peak toward lower binding energy. This shift indicates weakening of the Pb-I bonding in the [PbI₆]^4- framework^[18,60], likely producing a defect-rich buried interface populated by I- and Pb-related defects^[61]. In contrast, post-exposure annealing drove the two samples in opposite directions. In the control film, the post-exposure annealing further accelerated FAPbI₃ phase degradation. A new component appeared at ~620.1 eV, assigned to iodine in a less electron-rich environment than PbI₂ near vacancy and defect states^[61-63], and PbI₂ content increased at the expense of FAPbI₃. The thermal release of volatile iodine-related species generates oxidized iodine and modified iodine coordination environments (Eqs. 3-6)^[11,57,64]. In the target film, the trend was reversed. The FAPbI₃ phase partially recovered while the PbI₂ content decreased. Even when the FAPbI₃ lattice is weakened and partially decomposed by moisture, residual FA⁺ species retained within the film recombine with PbI₂ during the post-exposure annealing to restore FAPbI₃ (Eq. 7)^[65-67]. Supplementary Figure 15 shows a similar but weaker trend in the I 3d_5/2 spectra at a depth of 200 nm, further confirming that the buried interface is the dominant site of moisture-induced degradation. Figure 5C and D show the O 1s spectra deconvoluted into components corresponding to different chemical states. In the pristine films, the dominant component at 530.21 eV is assigned to SnO₂, with additional components from SnO at 529.49 eV, and Sn-OH/chemisorbed oxygen at 530.93 eV^[68]. The SnO and Sn-OH components reflect defect-related Sn²⁺ centers and surface hydroxylation at the SnO₂ layer^[69]. Upon humidity exposure [Figure 5C], the SnO₂ component shifted toward higher binding energy, accompanied by partial reduction of SnO₂ and a simultaneous increase in the SnO component. These changes indicate that moisture accumulation at the buried interface alters the local chemical environments within SnO₂ and increases the population of defect-related Sn²⁺ sites. Post-exposure annealing of the humidity-exposed control film promoted a new reaction without inducing reconversion into α-phase. At 170 °C, the interfacial oxygen became mobile and reacted with uncoordinated Pb at the buried interface to form Pb-O-related species (Eqs. 4-6). This interfacial oxygen comprises residual H₂O and -OH groups trapped at the interface together with reactive surface oxygen from the SnO₂ layer^[65,70,71]. Since the annealing was performed in ambient air, atmospheric oxygen can additionally oxidize mobile iodide, providing a parallel route to the same oxidized products. The co-presence of oxygen and moisture is known to aggravate the degradation of FA- and I-based perovskites, where surface I^- is oxidized to IO₃^-, which subsequently forms Pb(IO₃)₂-related products (Eq. 6)^[11,72,73]. Such iodate species are consistent with the oxidized iodine component resolved in Figure 5A, and their formation ultimately prevents recovery of the original Pb-I framework^[11]. The resulting Pb-O-related species are stable under ambient conditions, and barely reverse to Pb-I^[61], indicating that the post-exposure annealing of the humidity-exposed control film drives an irreversible decomposition of FAPbI₃. By contrast, the target film suppressed the formation of Pb-O-related species, and the SnO₂ binding energy remained constant across all stages. This indicates that the CsCl interlayer passivates the SnO₂ surface and blocks the irreversible reaction that produces Pb-O-related species during post-exposure annealing, thereby preserving the pathways for reconversion to the α-phase from PbI₂/FAI (Eq. 7) and from δ-phase (Eq. 8)^[74,75].

[Humidity exposure (H₂O accumulation at the SnO₂ surface)]

(1) $$ \alpha F A P b I_{3} \xrightarrow[{H}_{2} \mathrm{O}]{ } \delta F A P b I_{3} $$

(2) $$ F A P b I_{3} \underset{{H}_{2} \mathrm{O}}{\longrightarrow} P b I_{2}+F A I_{(a q)} $$

[Post-exposure annealing under a locally H₂O-rich environment (as in the control film)]

(3) $$ F A I+2 {H}_{2} {O} \underset{\Delta}{\rightarrow} H C O O H+2 N H_{3}+H I $$

(4) $$ P b I_{2}+{H}_{2} O \underset{\Delta}{\rightarrow} P b(O H) I+H I $$

(5) $$ 2 P b I_{2}+2 {H}_{2} O+{O}_{2} \underset{\Delta}{\rightarrow} 2 {Pb}(O H)_{2}+2 I_{2} $$

(6) $$ P b I_{2}+3 O_{2} \underset{\Delta}{\rightarrow} P b\left(I O_{3}\right)_{2} $$

[Post-exposure annealing under a locally H₂O-poor environment (as in the target film)]

(7) $$ P b I_{2}+F A I_{(a q)} \underset{\Delta}{\rightarrow} F A P b I_{3} $$

(8) $$ \delta F A P b I_{3} \underset{\Delta}{\rightarrow} \alpha F A P b I_{3} $$

Supplementary Figure 16 shows the Pb 4f_7/2 spectrum at a depth of 400 nm, near the buried interface, and Supplementary Figure 17 shows the corresponding spectrum at a depth of 200 nm, within the bulk of the perovskite film. In both spectra, an additional peak appears at higher binding energy than the PbI₂ component (~139.1 eV)^[76,77]. Pb-O-related species such as Pb(OH)₂, PbCO₃, and Pb(IO₃)₂ are known to form during degradation of the analogous perovskite system in conjunction with metal oxide layers^[70,71,78], and typically appear at higher binding energy than PbI₂, supporting the assignment of the new O 1s peak in Figure 5C to Pb-O-related species^[64,79]. Supplementary Figure 18 accordingly summarizes the depth-dependent XPS profiles of the Pb 4f_7/2 region, in which a substantial shift in binding energy appears only in the control sample after the post-exposure annealing and grows progressively on approaching the buried interface at greater depth. This again points to the buried interface, rather than the bulk perovskite, as the main driver of the irreversible reaction during the post-exposure annealing. Overall, deconvolution of the Pb 4f_7/2, I 3d_5/2, and O 1s spectra near the buried interface in Supplementary Figure 19 provides quantitative support for the mechanism proposed above. In the control sample, the FAPbI₃-related components declined as the PbI₂ components grew during the humidity exposure, and the post-exposure annealing converted the PbI₂ together with the Sn-OH/chemisorbed oxygen into Pb-O-related species, consistent across all three core levels [Supplementary Figure 19A-C]. The portion of Sn-OH/chemisorbed oxygen in O 1s is higher for the control film than for the target film throughout the series, from the pristine film to the humidity-exposed film and the post-exposure annealed film [Supplementary Figure 19C and F], supporting the water-blocking effect of the CsCl interlayer. The target sample followed neither pathway to completion, forming no Pb-O-related species and partially recovering its FAPbI₃ signal after the post-exposure annealing relative to the humidity-exposed state [Supplementary Figure 19D-F]. ToF-SIMS was additionally performed to cross-check the formation of Pb-O-related species with and without the CsCl interlayer. In Supplementary Figure 20, ToF-SIMS after post-exposure annealing revealed PbO^- fragments from both samples, with a markedly more intense signal near the buried interface in the control sample. The primary ion beam can itself generate PbO^- during sputtering^[80-82]. Because this artifact should be comparable for both samples, it cannot account for the excess observed in the control film. Pb-O-related species therefore formed extensively in the control film but not in the target film during post-exposure annealing. Furthermore, because the Pb-O products are typically amorphous, their formation is consistent with the absence of an additional crystalline phase in the XRD patterns in Figure 3^[78]. Consistently, Pb-O-related species increased appreciably only after post-exposure annealing, indicating that the annealing step, rather than the humidity exposure itself, is the kinetic trigger for the irreversible reaction. The elevated temperature accelerates iodide migration, Pb-I bond dissociation, and the transport of oxygen- and hydroxyl-related species^[83-85], thereby supplying both the undercoordinated Pb sites and the mobile reactants required for the reactions. The XPS results are consistent with the ToF-SIMS, XRD, and FT-IR trends, demonstrating the resilience afforded to FAPbI₃ by the CsCl interlayer at the buried interface, which suppresses the moisture- and thermally driven irreversible decomposition of FAPbI₃ into Pb-O-related species.

CsCl was introduced as an interlayer at the SnO₂/FAPbI₃ buried interface to probe humidity- and thermally driven degradation in n-i-p-structured PSCs. The CsCl interlayer passivated defects at the buried interface and modulated FAPbI₃ crystal growth, yielding higher photovoltaic performance with the largest gains in V_OC. Beyond the efficiency improvement, CsCl reinforced the structural resilience of FAPbI₃ against humidity-induced thermal degradation, a pathway initiated by moisture accumulation at the buried interface. Humidity exposure weakened the Pb-I framework through preferential H₂O interaction with FA⁺, generating the δ-phase and PbI₂. In the absence of the CsCl interlayer, subsequent thermal treatment then drove an irreversible reaction in which uncoordinated Pb reacted with surface oxygen at the buried interface to form Pb-O-related species. By contrast, the CsCl-treated interface suppressed Pb-O-related species formation and drove partial reconversion of the δ-phase and PbI₂ to the α-phase during post-exposure annealing. Beyond its interfacial role, the CsCl layer acted synergistically by releasing Cs⁺ into the perovskite lattice, where Cs⁺-[PbI₆]^4- coordination strengthens local bonding and suppresses structural distortion under environmental stress. In conclusion, the CsCl interlayer (i) promotes favorable FAPbI₃ crystallization in the pristine film, (ii) mitigates moisture-assisted disruption of the FA⁺-[PbI₆]^4- sublattice, and (iii) enables partial reconversion of the moisture-induced δ-phase and PbI₂ back to the α-phase while blocking the irreversible degradation pathway to Pb-O-related species at the SnO₂/FAPbI₃ buried interface during post-exposure annealing. These findings establish interface-engineered structural resilience as a practical design strategy for FAPbI₃-based PSCs.