Moisture-responsive crystallization strategy for efficient CsPbI₃ solar cells fabricated under high-humidity conditions

Abstract

The intrinsic phase instability of CsPbI₃ perovskites necessitates stringent fabrication conditions, significantly hindering the practical deployment. In the DMA-mediated CsPbI₃ nucleation system, the Cs⁺/DMA⁺ ion exchange critically governs the resulting film quality. Here, we employ a moisture-responsive crystallization strategy utilizing propyltriethoxysilane (PTES) to deposite CsPbI₃ under ambient air with high humidity (55%). We demonstrate that the siloxane groups can capture DMA⁺ in the intermediate DMAPbI₃, facilitating DMA⁺ extraction and Cs⁺ incorporation, thereby accelerating crystallization kinetics. This approach enables CsPbI₃ PSCs to achieve a power conversion efficiency (PCE) of 21.00% with an impressive fill factor (FF) of 86.1% while processing perovskite under relative humidity (RH) of 55%. Higher PCEs of 21.85% and 22.60% (certified 22.02%) were achieved for devices fabricated at a lower RH of 25% and for films spin-coated under an N₂ atmosphere followed by annealing in ambient air, respectively. Furthermore, PTES-treated devices exhibit excellent operational stability under ambient conditions.

CsPbI₃ perovskites face severe phase instability, limiting practical fabrication. The authors introduce a PTES-based moisture-responsive strategy enabling ambient processing and achieving up to 22.6% efficiency with improved stability.

Introduction

CsPbI₃ has garnered significant attention owing to its exceptional chemical stability and remarkable photoelectric performance^1–7. Recent advancements in CsPbI₃-based perovskite solar cells (PSCs) have demonstrated significant progress, achieving efficiencies over 22%⁸. Unfortunately, the small ionic radius of Cs⁺ ion leads to great lattice distortion⁹, rendering CsPbI₃ prone to a phase transition into the optically inactive δ phase in ambient air^10,11. This phase transition is significantly accelerated under the erosive effect of moisture¹². During film formation, the high solvation enthalpy of halide ions lowers the formation energy of vacancies, enabling water molecules to increase halide vacancy density. The vacancies weaken the ionic bonding between adjacent [PbI₆]⁴⁻ octahedra, catalyzing phase transitions^2,13,14. Additionally, moisture molecules can extract Cs⁺ ions from the surface lattice, further reducing the energy barrier for phase transitions driven by octahedral distortions^12,15. Consequently, most high-efficiency CsPbI₃ devices are fabricated within inert gases or low humidity (RH < 40%) environments to suppress degradation^16–20. This stringent environmental control presents a significant barrier to scalable, cost-effective production. Therefore, the fabrication of high-efficiency CsPbI₃ PSCs in ambient air without restrictions of relative humidity (RH) remains a significant challenge.

In recent years, organic molecule additive engineering has emerged as a widely adopted strategy to regulate the growth of CsPbI₃ thin films in air and maintain their phase stability^21,22. Reported additives generally function by interacting with atmospheric moisture to mitigate water-induced lattice degradation²³ or by introducing hydrophobic components that protect the perovskite during film formation²⁴. Organosilane and siloxane compounds such as 3-Aminopropyltrimethoxysilane (APTMS) have been used in hybrid perovskites to passivate grain boundaries and improve hydrophobicity^25–27. However, these treatments are applied mainly at the surface of MA/FA-based perovskites and rely on moisture-sensitive groups, which limit their effectiveness under the high-humidity conditions required for CsPbI₃ processing. And these approaches remain limited in the ability to regulate the crystallization dynamics of CsPbI₃ under ambient conditions.

To date, the crystallization of high-quality CsPbI₃ primarily relies on spontaneous, additive-free nucleation driven by DMA⁺/Cs⁺ ion exchange^5,28–30. Accelerating DMA⁺ removal while maintaining sufficient time for crystal growth is essential for forming well-defined γ-CsPbI₃ films^31–33. Previous studies further show that moderate humidity facilitates DMA⁺ extraction, whereas excessive moisture disrupts nucleation and leads to poor crystallinity^21,34,35. Thus, effectively harnessing atmospheric moisture to regulate CsPbI₃ crystallization is key to fabricating high-performance devices under high-humidity conditions. However, most reported crystallization-control strategies rely on hygroscopic functional groups (e.g., –NH₂, –COOH, or –SO₃⁻), whose inherent moisture sensitivity limits the applicability at elevated humidity^16,36. This gap underscores the urgent need for new crystallization-regulation mechanisms that remain robust under elevated humidity levels.

Here, we propose a moisture-driven crystallization regulation strategy for the preparation of high-quality CsPbI3 films under high-humidity conditions. As a multifunctional additive, propyltriethoxysilane (PTES) promotes the formation of dense CsPbI₃ films with enlarged grain sizes. We further reveal a direct interaction between siloxane species and DMA⁺, which plays a decisive role in guiding CsPbI₃ crystallization in humid environments. Namely, PTES accelerates the detachment of DMA⁺ ions from the DMAPbI₃ structure, facilitating Cs⁺ insertion into the [PbI₆]⁴⁻ framework. Moreover, PTES reacts with ambient moisture to form Si−O−Si and Si−O−Pb cross-linked networks that reinforce the lattice and stabilize the γ-phase even under elevated humidity. Notably, this crystallization-regulation mechanism arises intrinsically from the siloxane chemistry of PTES itself without assistance from additional functional groups. The resulting Si−O−Pb bonding minimizes internal defects, reduces iodine vacancies, and improves band alignment, enabling ultrafast charge transfer. Hence, PTES-treated CsPbI₃ PSCs achieved a record efficiency of 21.00% with an impressive fill factor (FF) of 86.1%. The strategy enables CsPbI₃ solar cells to be fabricated entirely in ambient air, with performance comparable to glovebox-processed devices, effectively overcoming environmental constraints in fabrication process.

Results

Organosiloxanes possess distinctive hydrolysis properties, allowing them to react with moisture and self-assemble into hydrophobic Si−O−Si networks^37,38. We employ propyltriethoxysilane (PTES) as an additive to regulate the growth of CsPbI₃ under a humid environment. Under moisture exposure, PTES undergoes hydrolysis to generate Si−OH groups, which spontaneously condense into a self-assembled Si−O−Si network (Supplementary Fig. 1)^39–41. Intriguingly, hydrolyzed intermediate Si−OH can coordinate with Pb²⁺ via Lewis acid–base interactions, forming Si−O−Pb bonds^40–42. This strong coordination anchors the [PbI₆]⁴⁻ octahedra and cross-links the CsPbI₃ lattice, thereby regulating the film quality and stabilizing the γ-phase^21,24,37. Furthermore, the trace water generated from PTES hydrolysis and condensation might even promote the crystallization of CsPbI₃^32,43. Although Trimethoxypropylsilane (PTMS) and 3-Aminopropyltriethoxysilane (APTES) were also evaluated, our comparison of their electrostatic potentials and their influence on CsPbI₃ film morphology (Supplementary Fig. 2 and Supplementary Note 1) highlights the clear advantage of PTES. PTES exhibits a milder hydrolysis process, which supports uniform crystallization of CsPbI₃ under humid conditions. In contrast, the faster hydrolysis of PTMS and the strong moisture-induced cross-linking of APTES lead to premature siloxane aggregation and severe film disruption. These differences indicate that PTES is the most suitable silane additive for stabilizing CsPbI₃ growth in high-humidity environments. Leveraging this chemical reactivity, PTES is introduced into the CsPbI₃ precursor solution to modulate crystallization dynamics and improve film formation under high-humidity conditions.

All CsPbI₃ films were fabricated via a one-step spin-coating method using ethyl acetate as the anti-solvent, followed by thermal annealing at 180 °C. All processes of perovskite films were conducted under ambient conditions with a relative humidity (RH) of ~55%. The untreated and PTES-treated CsPbI₃ films are labeled as the control and target samples, respectively.

To elucidate the PTES-assisted crystallization mechanism of CsPbI₃ under high-humidity conditions, we investigated the evolution of film crystallinity during annealing. As shown in Fig. 1a, the target films darken more rapidly than the control, indicating accelerated crystallization dynamics. XRD measurements of films annealed for different times (Fig. 1b, c and Supplementary Fig. 3) show the DMAPbI₃ peak (~11.8°) steadily decreases and disappears in target samples. In contrast, the control samples first show an increase, then a slow decline (Supplementary Fig. 4). This suggests that in control samples, DMA⁺ ions remain bound to the [PbI₆]⁴⁻ framework due to their higher binding energy, promoting intermediate DMAPbI₃ formation before eventual conversion to γ-CsPbI₃ via Cs⁺/DMA⁺ ion exchange during extended annealing³². However, PTES facilitates the early release of DMA⁺ through adsorption between siloxane groups and DMA⁺ in DMAPbI₃^40,44,45, suppressing intermediate phase crystallization and accelerating Cs⁺ incorporation. Furthermore, PTES further suppresses the crystallization of the DMAPbI₃ intermediate, facilitating additional lattice compression and enhancing the structural stability of CsPbI₃ (Supplementary Figs. 5 and 6 and Supplementary Notes 2 and 3).

Fig. 1. Crystallization dynamics characterization of perovskite films.

a Images of control and target samples at various annealing times in ~50% RH air. b, c XRD patterns of (b) control and (c) target samples annealed at different times. (In the color bar, bright yellow indicates higher intensity, while dark purple indicates lower intensity.) d, e Raman spectrum of (d) control and (e) target samples annealed at different times. (The red shading represents the asymmetric pattern of Pb-I. The blue shading indicates I₂ and I⁻ vacancies.) f, g Raman mapping spectra of (f) control and (g) target samples at 108 cm⁻¹. h Adsorption configurations and energies of PTES with DMA⁺, Pb²⁺, I⁻ and Cs⁺, i Corresponding adsorption energies for PTES on the four evaluated sites. j DMA⁺ escape energy barriers on different DMAPbI₃ facets with PTES. k Corresponding escape energy barriers for DMAPbI₃ on different facets.

To further elucidate the regulatory mechanism of PTES on the crystallization dynamics and lattice evolution of CsPbI₃, in situ Raman spectroscopy was systematically conducted during the annealing process (Fig. 1d). The characteristic peaks observed at 108 cm⁻¹ and 90 cm⁻¹ are attributed to the symmetric stretching vibration of Pb−I−Pb bridging bonds and the asymmetric breathing mode of [PbI₆]⁴⁻ octahedra, respectively. Upon thermal annealing, the thermally activated atomic rearrangement drives lattice relaxation and suppresses octahedral distortions, manifesting as a progressive redshift of the 108 cm⁻¹ mode. Concurrently, enhanced crystallinity suppresses defect-mediated localized lattice distortions, leading to the complete disappearance of the 90 cm⁻¹ feature. Notably, target samples showed a faster decrease of the 90 cm⁻¹ mode, indicating an accelerated transition to symmetric Pb–I vibrations. The Raman band at ~160 cm⁻¹ is attributed to I–I stretching vibrations of molecular iodine (I₂) or localized iodine aggregation near vacancy sites, originating from oxidation of I⁻ under ambient O₂/H₂O exposure. This process leads to the formation of iodine vacancies and lattice instability in the control film. In contrast, PTES incorporation suppresses this reaction by forming a hydrophobic and crosslinked Si–O–Si/Si–O–Pb network that inhibits moisture adsorption and stabilizes the Pb–I framework. Moreover, the regulation of CsPbI₃ crystallization by PTES is further supported by confocal Raman mapping of the 108 cm⁻¹ vibrational mode (Fig. 1e and f). Unlike the spatially heterogeneous signal in the control, target films display a more uniform and intense distribution, consistent with improved morphology in Fig. 1a. This enhancement results from homogeneous nucleation driven by Si–OH groups from PTES. These groups likely attach to precursor colloids, creating evenly distributed nucleophilic sites that lower the nucleation barrier and speed up CsPbI₃ crystallization.

We calculated adsorption energies of siloxane species on various atomic sites within the DMAPbI₃ intermediate (Fig. 1g). PTES shows the strongest adsorption at the DMA⁺ site (E_ad = –1.06 eV), followed by the I (–0.65 eV) and Pb (–0.54 eV) sites, indicating a preferential interaction that destabilizes DMA⁺. Building on these adsorption insights, we next evaluated the escape energy barriers of DMA⁺ on three representative facets of DMAPbI₃ (Fig. 1h and Supplementary Fig. 7). In the absence of PTES, DMA⁺ removal is hindered by positive activation barriers. By contrast, PTES significantly lowers the escape barriers across all facets, with the most pronounced reduction observed on the (001) plane, where the barrier drops from +0.32 eV to –0.58 eV. This facet-dependent modulation demonstrates that PTES reshapes the thermodynamic landscape of DMA⁺ volatilization, rendering its extraction energetically favorable and thereby accelerating the phase transition from DMAPbI₃ to CsPbI₃.

The accelerated crystallization dynamics may stem from adsorption between PTES and DMA⁺, which facilitates DMA⁺ extraction and Cs⁺ incorporation. These reactions were confirmed by ¹H nuclear magnetic resonance (NMR) and Fourier-transform infrared (FTIR) spectroscopy (Fig. 2a, b). In the NMR spectra, the disappearance of Si−O − C₂H₅ peaks at 3.73 and 1.14 ppm in PTES upon mixing with DMAPbI₃ indicates siloxane group reactivity. Concurrently, the −CH₃ peak of DMAPbI₃ shifts from 2.55 ppm, confirming changes in the chemical environment of DMA⁺. Detailed ¹H spectra are provided in Supplementary Figs. 8 and 9. FTIR analysis further corroborates the reaction. The −NH⁺− stretching vibration in DMAPbI₃ shifts from 1562 cm⁻¹ to 1570 cm⁻¹ after PTES treatment, while the Pb−N stretching at 805 cm⁻¹ disappears, replaced by the appearance of Si−N band at 936 cm⁻¹. These spectral transitions indicate the breaking of Pb−N bonds and the formation of Si−N bonds under thermal and humid conditions, accelerating the detachment of DMA⁺ and the insertion of Cs⁺ into [PbI₆]⁴⁻. Simultaneously, on the PTES-treated DMAPbI₃ samples, the disappearance of Si−O − C signals at 950 cm⁻¹ and 775 cm⁻¹, along with the appearance of Si−OH (842 cm⁻¹) and Si−O−Si (1017 cm⁻¹), further confirms the hydrolysis of PTES and the subsequent formation of cross-linked networks. To demonstrate that these reactions during film formation, FTIR was conducted on control and target samples annealed at 70 °C and 180 °C (Fig. 2c). At 70 °C pre-annealing of the target films, the appearance of Si−N (935 cm⁻¹) and Si−OH (842 cm⁻¹) peaks indicates early-stage adsorption and hydrolysis. Upon further annealing, Si−N bonds decompose while Si−OH species condense into Si−O−Si or react with Pb to form Si−O−Pb networks. These cross-linked networks not only stabilize the perovskite lattice but also inhibit moisture penetration, thereby preserving the γ-CsPbI₃ phase during film formation^21,24.

Fig. 2. Characterization of interaction between PTES and perovskite.

a The NMR ¹H and b FTIR spectra of pure DMAPbI_3, pure PTES, and DMAPbI₃ treated with PTES samples. c, The FTIR spectra of control and target films under 70˚C and 180˚C annealing. d Pb 4 f and e Si 2 P XPS spectra of control and target films. f Mechanism schematic of PTES-assisted high-quality crystallization of CsPbI₃ under high humidity conditions.

X-ray photoelectron spectroscopy (XPS) was employed to investigate the surface composition of CsPbI₃ films and further elucidate the role of PTES under high-humidity conditions. Figure 2d depicts that the Si 2p signal at ~102 eV appears exclusively in target film, confirming the retention of PTES within the CsPbI₃ matrix. Moreover, time-of-flight secondary ion mass spectrometry (ToF–SIMS) depth further confirms that hydrolyzed PTES is uniformly incorporated into the CsPbI₃ bulk in the form of Si–O–based species, rather than being confined to the interface (Supplementary Fig. 10). Meanwhile, the N 1s signal (Supplementary Fig. 11a) is nearly absent in the target sample but persists in the control, indicating efficient DMA⁺ removal facilitated by PTES, consistent with the XRD results. The Pb 4f core-level spectra (Fig. 2e) reveal peak positions at 142.4 and 137.6 eV for the target film, corresponding to a 0.3 and 0.2 eV shift, respectively, compared to the control. The peak shift suggests an interaction between PTES and Pb, attributed to the formation of Si−O−Pb cross-linking networks. In addition, a slight shift of the I 3d peak toward lower binding energy in the target film (Supplementary Fig. 11b) suggests a reduced density of iodine vacancies. These findings demonstrate that PTES effectively suppresses halide vacancy formation and enhances film quality by stabilizing the CsPbI₃ surface chemistry under humid ambient conditions.

Mechanistically, hydrolyzed PTES produces Si–OH and Si–O– species that interact with DMA⁺ and regulate its release from the DMAPbI₃ intermediate, enabling a gradual transition toward CsPbI₃. As annealing proceeds, PTES condenses to form Si–O–Si and Si–O–Pb linkages, thereby strengthening local lattice connectivity and mitigating structural fluctuations during phase conversion. This coordinated process accelerates Cs⁺ incorporation into the [PbI₆]⁴⁻ framework, promotes uniform crystallization, and suppresses defect formation. At the same time, the in situ-formed hydrophobic siloxane network reduces moisture ingress, thereby supporting the stabilization of γ-CsPbI₃ under humid conditions. Together, these effects enable PTES to harness ambient moisture constructively and yield high-quality CsPbI₃ films even in high-humidity environments.

X-ray diffraction (XRD) analysis (Supplementary Fig. 12 and Supplementary Note 4) reveals characteristic peaks at 14.4° and 28.9°, corresponding to the (110) and (220) planes of γ-phase CsPbI₃^46,47. A minor peak at 11.8° attributed to DMAPbI₃ is present in control sample, indicating incomplete evaporation of DMA⁺³². The absence of DMAPbI₃ peak in the target sample suggests that PTES enhances DMA evaporation, promoting complete ions conversion. Particularly, the (110) peak in the target sample exhibits a rightward shift and reduced shoulder intensity, implying improved preferential orientation, lattice compression, and reduced interplanar spacing—factors associated with enhanced structural stability⁴⁸. In addition, the markedly smaller slope in the W–H plot further confirms the release of internal microstrain within the perovskite lattice. To further investigate the lattice stress-strain behavior, grazing incidence X-ray diffraction (GIXRD) was performed at varying incident angles (ω) (Fig.3a and b). In control samples, peak shifts toward higher angles with increasing ω indicate surface tensile strain. In contrast, the target samples show negligible peak movement, confirming strain relaxation and improved lattice uniformity⁴⁹. The specific mechanism of stress-strain release and the microstrain values are depicted in Fig. 3c.

Fig. 3. Characterization of the crystallographic and photovoltaic properties of perovskite films.

a, b GIXRD patterns with different incident angle (ω) of (a) control and (b) target perovskite films. c, The microstrain values and stress release mechanism diagram of perovskite film. d–f KPFM of (d) control and (e) target perovskite films, and (f) corresponding surface potentials. g PL, h TRPL, and i PL mapping spectra of control and target samples on the glass substrates.

The morphology of the control and different concentration target samples was examined utilizing top-view scanning electron microscopy (SEM) (Supplementary Fig. 13). The control films exhibit a non-compact morphology with visible pinholes and cracks, primarily resulting from moisture-induced degradation, in agreement with previous studies^21,22,24,35. Following PTES treatment, the CsPbI₃ films exhibit significantly improved compactness, highlighting the effectiveness of PTES in modulating film quality under high-humidity conditions. Cross-sectional SEM images (Supplementary Fig. 14) further reveal enhanced crystallinity and a more ordered structure in the target samples. Atomic force microscopy (AFM) analysis (Supplementary Fig. 15) shows a marked reduction in surface roughness from 17.10 nm (control) to 7.83 nm (target), indicating a smoother film surface and improved crystallinity, which are favorable for enhanced interfacial contact.

As shown in the UV-Vis spectra (Supplementary Fig. 16 and Supplementary note 5), both perovskite films display comparable absorption profiles, with the target film exhibiting enhanced absorption near 550 nm, attributed to improved crystallinity and film quality. Ultraviolet photoelectron spectroscopy (UPS) (Supplementary Fig. 17a) reveals valence band maxima of −5.47 eV (control) and −5.27 eV (target), yielding conduction band minima of −3.76 and −3.56 eV, respectively. The corresponding energy level diagram (Supplementary Fig. 17b) illustrates that PTES induces an upward shift in energy levels, improving alignment with the hole transport layer and facilitating hole extraction. Notably, the Fermi level of the target film shifts upward by ~0.1 eV, enhancing the driving force for carrier separation and reducing open-circuit voltage (V_oc) losses. KPFM analysis (Fig. 3d–f) further confirms this shift, showing a higher and more uniform contact potential difference (CPD) in the target film. The influence of PTES on carrier dynamics was examined via steady-state and time-resolved photoluminescence (PL and TRPL) measurements. As shown in Fig. 3g, the target CsPbI₃ film exhibits markedly enhanced PL intensity with a blue-shifted emission peak, indicative of reduced defect density. TRPL analysis (Fig. 3h and Supplementary Table 1) reveals a substantially prolonged average carrier lifetime from 7.90 ns (control) to 44.25 ns (target), confirming suppressed nonradiative recombination. Furthermore, confocal PL mapping (Fig. 3i and Supplementary Fig. 18) shows a more intense and spatially uniform emission from the target film, highlighting improved film homogeneity and compositional uniformity.

The J-V characteristics with FTO/TiO₂/SA/CsPbI₃/PTABr/Spiro-OMeTAD/Au, photovoltaic parameters and cross-sectional SEM image of target device are shown in Fig. 4a and Supplementary Table 2. The control devices achieve a PCE of 18.19%, whereas the target device achieved the PCE of 21.00% with an outstanding FF of 86.1%. It is the highest reported FF among CsPbI₃ PSCs to date. The reduced hysteresis in the target device can be attributed to the PTES-induced reduction of interfacial and bulk defect densities, which suppresses ion migration and interfacial charge accumulation. The target device also demonstrated excellent repeatability, with the tighter PCE distribution, as supported by the statistical data from 40 devices (Fig. 4b and Supplementary Fig. 19). Optimization of PTES concentration identified 4 mol% (relative to DMAPbI₃) as optimal (Supplementary Figs. 20, 21 and Table 3). The high efficiency and strong reproducibility also benefit from the commonly used additives DMACl and PbAc₂. XRD, PL, and UV–vis measurements (Supplementary Fig. 22 and Supplementary Note 6) confirm their respective roles in both control and PTES-treated samples, and demonstrate that the additional DMA and Pb sources do not interfere with PTES-mediated crystallization or introduce extra DMA⁺ or Pb²⁺ into the final CsPbI₃ films. Figure 4c displays the external quantum efficiency (EQE) curves and corresponding integrated current densities of the control and target devices, yielding 18.97 mA/cm² and 19.79 mA/cm², respectively. The steady-state power output (SPO), tracked under maximum power point (MPP) conditions (Fig. 4d), further confirms device stability. The target device maintains a current density of 19.48 mA/cm² and PCE of 20.65%, compared to 18.47 mA/cm² and 17.64% for the control.

Fig. 4. Photovoltaic performance of control and target PSCs.

a J-V curves of control and target devices with the structure FTO/TiO₂/SA/CsPbI₃/PTABr/Spiro-OMeTAD/Au (inset: cross-sectional SEM image of target device). b Statistical graphs of the control and target devices (20 devices of each sample were fabricated for each condition). c EQE curves and integrated current curves. d SPO curves of various devices. e, f J-V curves of devices fabricated under (e) RH ≈ 25% and (f) N₂ atmosphere. g, h Reported PCE statistics of CsPbI₃ PSCs fabricated under the (g) different humidity and (h) N₂ atmosphere. (The shaded area indicates high humidity). i MPP tracking of the control and target devices under continuous illumination in N₂ atmosphere. j, Steady power output of the control and target devices at 1.05 V under ambient humidity of 50% (3 min/point).

To further verify the scalability of the PTES approach, we fabricated 1 cm² devices (Supplementary Fig. 23), which exhibited a PCE of 18.81%, confirming the applicability of this method for larger-area perovskite solar cells. Moreover, to validate the general applicability of PTES, devices were fabricated under varying humidity levels (Fig. 4e, f and Supplementary Fig. 24). The target device achieved an impressive PCE of 22.60% (prepared in an N₂ atmosphere and annealed in air), with a certified efficiency of 22.02% (Supplementary Fig. 25), as well as 21.85% under RH ~ 25%. The statistical chart of target devices is shown in Supplementary Fig. 26. PCE evolution under different humidity conditions demonstrates the excellent reproducibility and highlights the universal efficacy of PTES across environments (Supplementary Fig. 27). Notably, the target devices consistently outperform prior CsPbI₃ PSCs fabricated under similar conditions, with substantial enhancements in FF as illustrated in Fig. 4g, h and Supplementary Table 5. Furthermore, we performed XPS measurements on target CsPbI₃ films prepared under different humidity conditions (RH = 25%, 50%, 75%) to evaluate the effect of moisture on siloxane crosslinking (Supplementary Fig. 28, Supplementary Note 7 and Table 6). The quantitative XPS results show that humidity strongly regulates PTES hydrolysis and crosslinking. As RH increases from 25% to 50%, the Si and O signals rise markedly, indicating enhanced formation of the Si–O–Si network. At RH = 75%, the crosslinking level approaches saturation. Device performance follows this trend, indicating that siloxane crosslinking is intrinsically beneficial and enhances moisture tolerance. Nevertheless, excessive humidity still induces structural degradation even at higher crosslinking levels. The markedly faster efficiency loss observed in control devices under high humidity further highlights the protective role of siloxane crosslinking. Moreover, even for films processed in a N₂ atmosphere, PTES-mediated crystallization still occurs because annealing is conducted in ambient air with RH of 20–30%, allowing sufficient hydrolysis and crosslinking to influence CsPbI₃ nucleation and growth.

Space charge limited current (SCLC) measurements were conducted to quantify trap densities in the perovskite films (Supplementary Fig. 29 and Supplementary Note 8), highlighting the effectiveness of PTES in suppressing defect formation. Additionally, SCLC fitting reveals that the target device exhibits a significantly improved electron mobility (Supplementary Fig. 30 and Supplementary Note 9). Dark J–V characteristics (Supplementary Fig. 31) revealed reduced leakage current in the target device, indicating suppressed carrier recombination due to fewer trap states. Mott–Schottky analysis further showed an increased built-in potential (V_bi) of 1.04 V in the target device, reflecting improved charge separation efficiency, consistent with the observed Fermi level shift (Supplementary Fig. 32 and Supplementary Note 10). Electrochemical impedance spectroscopy (EIS) analysis (Supplementary Fig. 33 and Table 7) demonstrated lower series resistance (R_s) and charge-transfer resistance (R_ct) in target devices, corroborating the enhanced charge transport and extraction properties imparted by the additive.

To assess the impact of PTES on the long-term stability of PSCs, various storage stability tests were conducted on both control and target devices. Evidently, the presence of PTES significantly enhances the long-term stability of the target devices (Supplementary Figs. 34, 35, Supplementary Notes 11, 12 and Table 8). MPP tracking under varying conditions further corroborated the improved stability (Fig. 4i, j). Under N₂ atmosphere, the encapsulated target device retained 90% of its initial PCE after 800 h of continuous illumination. In ambient air at 50% RH, the unencapsulated target device maintained ~80% of its efficiency after nearly 100 h, significantly outperforming the control. These results highlight the efficacy of the PTES additive strategy in enhancing both intrinsic and extrinsic stability of CsPbI₃ perovskite solar cells, even under challenging environmental conditions.

Discussion

In summary, PTES reacts with ambient moisture to form a hydrophobic network through Si−O−Pb and Si−O−Si bonding, leading to in situ crosslinking of the perovskite lattice that strengthens lattice interactions and effectively prevents further moisture intrusion. Additionally, the PTES promotes the separation of DMA⁺ ions from the DMAPbI₃ framework to accelerate the crystallization dynamics of CsPbI₃, rendering high-quality and phase-stable CsPbI₃ films. Consequently, champion PCEs of 21.00% and 21.85% were obtained under ambient conditions at 55% RH and 25% RH, respectively, while a higher PCE of 22.60% was achieved for devices fabricated by spin-coating in an N₂ atmosphere followed by annealing in ambient air. The minimal performance variation across different humidity levels underscores the broad ambient tolerance of this strategy, with the result at >50% RH representing a state-of-the-art efficiency for CsPbI₃ PSCs. Furthermore, the devices exhibit excellent long-term operational stability. This work offers a viable approach to overcoming environmental constraints in the fabrication of inorganic perovskites, paving the way for scalable, robust device development.

Methods

Materials

Titanium tetrachloride (TiCl₄, 99.9%), lead acetate (PbAc₂, 99%), Dimethylamine hydrochloride (DMACl, 99%), Triethoxypropylsilane (PTES, 98%), 3-Amino-1-propanesulfonic acid (SA, 97%), and Ethyl acetate (EA, 99.7%) were purchased from Aladdin. Lead(II) iodide (PbI₂, 99.99%), cesium iodide (CsI, 99.999%), Phenylethylamine Hydroiodide (PEAI, 99%), 4-tert-butylpyridine (TBP, 96%), 2,2’,7,7’-Tetrakis(N, N-di-p-methoxyphenylamine)−9,9-spirobifluorene (Spiro-OMeTAD, 99.86%) and tris (2-(1H-pyrazol-1-yl)−4-tert-butylpyridine) cobalt(III) tri[bis(trifluoromethane)sulfonimide] (FK209) were purchased from Advanced Election Technology Co., Ltd. N, N-Dimethylformamide (DMF, 99.8%), Dimethyl sulfoxide (DMSO, 99.9%), Chlorobenzene (CB, 99.8%), Li-bis-(trifluoromethanesulfonyl) imide (Li-TFSI, 99.5%) and Hydriodic acid (HI, 57 wt%), Phenyltrimethylammonium bromide (PTABr, 98%) were purchased from Sigma-Aldrich. And hypophosphorus acid (H₃PO₂, 50 wt% aqueous solution) was purchased from Macklin. Poly(3-hexylthiophene-2,5-diyl) (P3HT) were purchased from Xi’an Yuri Solar Co., Ltd. We did no further purification to all the experimental materials before we used.

Synthesis of DMAPbI₃

DMAPbI₃ [DMA = (CH₃)₂NH₂⁺] was synthesized using PbI₂ (2.305 g) dissolved in 5 mL DMF at 80 °C under active stirring for 30 min in air atmosphere. Immediately after, 5 mL HI were added to the solution, and a small amount of H₃PO₂ was subsequently added into the mixture to form a transparent solution. Stop stirring and allow it to cool to room temperature. After standing for two hours, collect the precipitate, wash it with ethanol 4-5 times, and then vacuum dry it overnight at 60 °C²⁹.

Preparation of precursor solution

For the preparation of control devices under air conditions, the perovskite precursor solution (1M) was prepared by dissolving 0.634g DMAPbI₃, 0.2676g CsI, 7.5mg PEAI, 2.5mg DMACl, and 10mg PbAc₂ in DMF/DMSO mixed solvent (volume ratio = 4:1), which was stirred for 16 h. For the target devices, PTES was added at 3–6% of the DMAPbI₃ molar amount, and all other conditions matched those of the control CsPbI₃ precursor. For the preparation of champion devices under N₂ atmosphere, the perovskite precursor solution (0.8 M) was prepared by dissolving 0.5072 g DMAPbI₃, 0.2141 g CsI, 7.5 mg PEAI, 2.5 mg DMACl, 10 mg PbAc₂ and 9.6 μL PTES in 1 mL DMF/DMSO mixed solvent (volume ratio = 4:1). Spiro-OMeTAD precursor solution was perpared by dissolving 90 mg of spiro-OMeTAD, 40 μL of 4-tBP, 23 μL of Li-TFSI solution (520 mg Li-TFSI in 1 mL acetonitrile) and 20 μL FK209 solution (300 mg FK209 in 1 mL acetonitrile) in 1 mL of chlorobenzene.

Device fabrication

Fluorine-doped tin oxide glass (FTO, 2*2 cm²) is cleaned for 30 min with detergent, deionized water, acetone, ethanol, and deionized water. Then, FTO is treated with oxygen plasma for 5 min. The TiO₂ layer was deposited by immersing FTO glass substrates in 200 mL aqueous solution containing 4.5 mL titanium tetrachloride at 70 °C for 60 min, then rinsed with distilled water and annealed at 200 °C for 30 min. Before spin-coating the perovskite precursor solution, spin-coat 0.5 mg/mL of SA aqueous solution at a speed of 4000 rpm for 20 s, and then anneal at 100 °C for 10 min. The perovskite precursor solution was spin-coated on the top of preheated compact TiO₂ substrate at 600 rpm for 6 s, 3000 rpm for the 30 s. At 15 seconds to the end of the spin coating, 300 μL of antisolvent ethyl acetate (EA) were dropped within 1 s. After that, the precursor film was annealed at 180 °C for 10 min in the air (RH 25-75%). 1 mg/ml of PTABr was applied to the upper surface of CsPbI₃ at 4000 rpm 20 s and annealed at 100°C for 10 min. Subsequently, the Spiro-OMeTAD hole transport layer (HTL) was prepared by dynamic spin coating at 1500 rpm for 30 s. The resulting films are then oxidized for 16–24 h. For stability test devices, 1 mg/ml PMMA CB solution is coated onto the perovskite film at 6000rpm for 20s, and then annealed at 100 °C for 10min. And HTL is replaced with 10 mg/ml P3HT, spin-coated at 3000 rpm for 30 s, and annealed at 100 °C for 30 min. Finally, a 75 nm thick Au layer was deposited by thermal evaporation.

Device fabrication in N₂ atmosphere

The perovskite precursor solution was coated onto the TiO₂ substrate modified with SA at 600 rpm 6 s and 3000 rpm 30 s. The substrate was then annealed in N₂ atmosphere at 70 °C until the film became transparent and a little foggy. Transfer to an environment with 25-35% RH and heat at 180 °C for 10 min. After completion, cool to room temperature and then transfer to N₂ atmosphere. The remaining steps are the same as those for air preparation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.H.L., Q.D.T. and J.B.Y. supervised the project. J.H.L. and W.D.R.D. were convinced of the idea. W.D.R.D. fabricated the CsPbI₃ PSCs and conducted most of the characterizations. Y.Z.G., J.Q.Z. and Z.X.P. were responsible for the SEM and participated in the optimization of CsPbI₃ PSCs. X.L.L. were responsible for the in situ XRD and Raman characterizations of CsPbI₃ film formation. H.J.H. provided support for device certification. C.C. contributed to the ToF-SIMS and XPS measurements and assisted with the provision of experimental conditions during the revision and response to reviewers. S.L.W. were responsible for the DFT simulation. X.B.W. and T.M. contributed to the analysis and provided advice. W.D.R.D. wrote the first draft, J.H.L., Q.D.T. and J.B.Y. revised the manuscript. All authors contributed to the proofs of the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All the data supporting the findings of this study are available within this article and its Supplementary Information. Any additional information can be obtained from corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69687-4.