Mechanistic insight into α-to-δ phase transition and stabilizing α-phase FAPbI3 via regulating δ-phase orientation Authors: Chuwu Xing, Yongzhe Li, Qinhui Bao, Miao He, Zhiqiang Lu, Yibin Zhang, Zhike Liu, Tianjin Zhang, Duofa Wang Published: 2026 ABSTRACT AbstractThe structural instability of formamidinium-based perovskite primarily originates from spontaneous α-to-δ phase transition. Although various suppression strategies have been proposed, developing a definitive solution remains challenging due to incomplete understanding of the physical mechanisms governing this phase transition. This work reveals that the coherent interface with tensile strain between perovskite α-phase (011) and δ-phase (110) planes acts as an active site for α-to-δ phase transition. Building on this insight, two-dimensional perovskitoid materials are introduced into the perovskite matrix, which strongly interact with the δ-phase (110) plane to form the 2D/δ interface. The large interfacial strain disrupts δ-phase orientation, suppresses the formation of α/δ transition interfaces, and thus enhances α-phase stability. This strategy enables FA-based perovskite solar cells to achieve an efficiency of 25.61%, with unencapsulated devices maintaining 90% initial efficiency after 1,000 h at 70% relative humidity. This work provides a mechanistic understanding and a universally applicable strategy to stabilize FA-based perovskites.Subject terms: Solar cells, Energy Introduction Formamidinium-based (FA-based) perovskite solar cells (PSCs) have demonstrated excellent photovoltaic performance due to their narrow bandgap (Eg) and superior thermal stability, achieving over 27% efficiency in single-junction cells1–3. However, their highly symmetric cubic α-phase (Pm−3m) structure exhibits thermodynamic metastability due to the oversized FA+ cation, yielding a Goldschmidt tolerance factor (t = 0.987) at the edge of stability interval (0.8-1.0)4–6. The thermodynamic metastability of FA-based perovskite drives spontaneous α-to-δ-phase transition at ambient conditions, forming a non-photoactive yellow δ-phase that degrades PSCs performance7–9. Previous stabilization strategies employed smaller cations (MA+, Cs+, Rb+) for partial FA+ substitution in α-phase FAPbI3, adjusting the tolerance factor10–13. However, the easily decomposable property of MA+ compromises both the device efficiency and operational stability14. In addition, Cs+ incorporation has emerged as a promising approach for enhancing the stability of FA-based PSC, accelerating commercial viability15,16. Despite the advantage of Cs+ incorporation, FACs-based PSCs still exhibit limited stability due to persistent moisture-triggered phase transitions17,18. To address the α-phase instability issue of FA-based perovskites, it is essential to explore and understand the physical mechanisms underlying their phase transitions, for developing effective strategies to inhibit the notorious phase transition. The metastability of the α-phase fundamentally originates from its intrinsic lattice softness, with lattice strain serves as a primary drive for phase transitions19,20. Thermal stress generated by thermal expansion coefficients mismatch between perovskite films and substrates compromises structural integrity through tensile strain, though such tensile strain can be effectively mitigated by optimizing annealing kinetics or introducing flexible organic interlayers21,22. Concurrently, the phase transition during perovskite formation induces lattice parameters evolving, leading to localized lattice strain. This lattice strain can be effectively suppressed through substitutional doping with isovalent/aliovalent ions at lattice sites23,24. Nevertheless, current research predominantly addresses strain engineering without starting from the basic phase transition mechanism to find a solution25–27. Our previous work has revealed that the significant lattice mismatch (~ 3%) between the α-phase and δ-phase generates tensile stress at the interface, which may act as a driving force to initiate the phase transition process28. Moreover, recent work demonstrates that defect states in FAPbI3 significantly reduce activation energy for α-to-δ phase transition, while the nuclei formation of δ-phase further lowers the transition barrier, thereby accelerating the degradation of α-phase29,30. This autocatalytic δ-phase induced phase transition may be the fundamental reason for the rapid phase transition of FACs-based perovskite under ambient conditions. Therefore, an in-depth analysis of the physical mechanisms of δ-phase mediated α-phase transitions, followed by the regulation of the phase transition pathway through rationally designed strategies, will become a key direction for enhancing α-phase perovskite stability. In this work, we identify and characterize the α-to-δ phase transition interface in FA0.9Cs0.1PbI3 perovskite, and reveal that the interfacial tensile strain drives preferential α-phase degradation to δ-phase. By incorporating a two-dimensional (2D) perovskitoid structurally analogous to δ-phase into α-phase perovskite, the strong 2D/δ phase interaction regulates δ-phase orientation, thereby suppressing the formation of α/δ phase transition interface and fundamentally addressing the instability issue of α-phase. The FACs PSCs based on this strategy achieve a power conversion efficiency (PCE) of 25.59%, with unencapsulated device maintaining 90% initial efficiency after 1000 h aging under 70% relative humidity (RH). Results Observation of α-to-δ phase transition interface This work employed mixed-cation lead halide perovskite FA0.9Cs0.1PbI3 (FACs) as the photoactive layer that was prepared via one-step spin-coating method. To investigate the phase transition mechanism of the FACs perovskite, the FACs perovskite films aging under 70% RH was analyzed through X-ray diffraction (XRD). As shown in Fig. 1a, the as-grown film exhibits a predominant cubic phase (α-phase). After 100 h aging, the (100) plane of the hexagonal phase (δ-phase) begins to appear near 11.8°, and the diffraction peak intensity significantly increases during aging. Notably, the diffraction peaks of other planes for δ-phase maintain minimal intensities, indicating that the preferential orientation of δ-phase along the [001] crystallographic direction. To quantitatively investigate the content and microstructural parameters of α-phase and δ-phase in the perovskite film, we conducted fitting calculation on the XRD results of the aged perovskite using the X-ray refinement method (the fitting diagram shown in Supplementary Fig. 1). The obtained contents of the α-phase and δ-phase, along with the microstructural parameters, are summarized in Supplementary Table 1. The content variation of δ-phase in FACs perovskite with aging time is illustrated in Fig. 1b, revealing that the content of δ-phase increases exponentially with prolonged aging time. This phenomenon indicates that the phase transition from α-phase to δ-phase accelerates as the δ-phase form, demonstrating that δ-phase significantly impacts the structural stability of α-phase negatively. To further elucidate the mechanism by which the δ-phase accelerates the phase transition of perovskite, we quantified microstrain through microstructural parameters analysis (computational methodology detailed in Supplementary Information). As illustrated in Fig. 1c, the aging process induces progressively increasing tensile strain in the α-phase lattice while the δ-phase demonstrates a corresponding release of compressive strain. This strain evolution correlates with the observed phase content redistribution, with a marked decrease in α-phase accompanied by δ-phase proliferation. In light of the variations in phase content, the overall strain of the α-phase and δ-phase can be defined as the product of the average microstrain and their respective phase fractions, as shown in Supplementary Fig. 2. During the aging process, the overall strain in both the α-phase and δ-phase exhibits a consistent trend, which increases initially and then decreases. Specifically, the overall strain reaches its maximum when the two phases are present in nearly equal amounts, and is minimal when the composition approaches a single-phase state. Therefore, we propose that these reciprocal strain patterns originate from interfacial interactions between α-phase and δ-phase at their phase boundaries. The exponential growth of δ-phase contents creates 1) a geometric amplification of α/δ interfacial areas; 2) progressive accumulation of α-phase lattice strain through constrained transformation; 3) concurrent strain relaxation in δ-phase due to diminishing mechanical constraints from the receding α-phase. This reinforced strain interaction establishes a mechanism that the interfacial strain both drives and is amplified by the α-to-δ phase transformation process. Observation of α-to-δ phase transition interface. a XRD pattern of the FACs perovskite after aging for 0 h, 100 h, 200 h, 300 h and 400 h. b δ-phase content of the aging process of FACs perovskite films obtained by XRD refinement. c Microstrain of the aging process of FACs perovskite α-phase and δ-phase obtained by XRD refinement. d HRTEM image of the out-of-plane α/δ interface in aged FACs perovskite, with α-phase lattice fringes highlighted in blue, δ-phase lattice fringes highlighted in red, and the out-of-plane α/δ interface denoted by a blue dashed line. The scale bar in the HRTEM image is 2 nm. e Schematic illustration of the out-of-plane α/δ interface structure, with the out-of-plane interface represented by a blue dashed box consistent with previous notation. f Octahedral spacing at the out-of-plane α/δ interface of α-phase and δ-phase obtained by XRD refinement. g HRTEM image of the in-plane α/δ interface in aged FACs perovskite, and the in-plane α/δ interface denoted by a red dashed line. The scale bar in the HRTEM image is 2 nm. h Schematic illustration of the in-plane α/δ interface structure, with the in-plane interface represented by a red dashed box. i Octahedral spacing at the in-plane α/δ interface of α-phase and δ-phase obtained by XRD refinement. In order to further intuitively and profoundly investigate the α/δ interface and its effect on phase transition in FACs perovskite, the coherent interface between the α and δ phases in the 300 h aged perovskite film was observed by the high-resolution transmission electron microscopy (HRTEM). Two distinct phase interfaces, namely out-of-plane and in-plane, were observed between the α and δ phases (Here, the out-of-plane interface is defined as the α/δ phase interface perpendicular to the substrate surface, while the in-plane interface is the α/δ phase interface parallel to the substrate surface). As shown in Fig. 1d, the out-of-plane coherent interface was demarcated by the blue dashed line and forms through epitaxial matching between the α-phase (011) plane (d = 4.4 Å) and δ-phase (110) plane (d = 4.3 Å). The fast Fourier transform (FFT) analysis enabled a more accurate assignment of the crystal planes (Supplementary Fig. 3). The FFT pattern for the α-phase region exhibits a standard cubic crystal diffraction arrangement. The measured distance from the diffraction spots to the central beam is 2.283 nm−1, corresponding to a lattice spacing of 4.4 Å, which matches the (011) plane of the α-phase. For the δ-phase, the distance from the diffraction spots to the central beam is 2.316 nm−1, corresponding to a spacing of 4.3 Å, and corresponding to the (110) plane. Crystal model (Fig. 1e) reveals interplanar distances of dα1 = 4.47 Å for the α-phase and dδ1 = 4.33 Å for the δ-phase that align with experimental measurements. Contrary to the tensile strain observed in α-phase (Fig. 1c), the smaller lattice of δ-phase at the interface will induce a compressive strain on α-phase, suggesting a minimal influence of α/δ out-of-plane interfaces on perovskite’s macroscopic strain evolution. In addition, the changes in dα1 and dδ1 at the α/δ out-of-plane interface were calculated through XRD refinement (Fig. 1f). The lattice mismatch was calculated, which slightly increased from the initial 2.95% to 3.12% (the calculation method provided in the Supplementary information). The subtle variation of lattice parameters further indicates that the strain effect generated at the α/δ out-of-plane interface is weak, associating with a highly oriented δ-phase. As depicted in Supplementary Fig. 4, the δ-phase has a one-dimensional (1D) structure with coplanar octahedrons, and the XRD pattern (Fig. 1a) also reveals that the δ-phase exhibits a strong preferred [001] orientation, so the δ-phase perovskite tends to grow along the crystallographic c-axis direction. The δ-phase (110) plane is parallel to the c-axis, having a 30° crystallographic angle with the (100) plane. Thus, the δ-phase (110) plane exhibits a similar [001] orientation characteristic along the c-axis as the (100) plane. Due to this high orientation of the δ-phase, the number of out-of-plane interfaces formed with the α-phase will be much less than that of in-plane interfaces. Consequently, the effect of out-of-plane interfaces on the stability of the α-phase is very limited. Figure 1g presents the HRTEM results at the α/δ in-plane interface that demarcated by the red dashed lines, showing that the spacing of quintuple octahedra in the α-phase is 15.6 Å and larger than the 15.2 Å in the bulk phase (Supplementary Fig. 5). The HRTEM result suggests that the α-phase near the α/δ in-plane interface experiences the δ-phase-imposed tensile strain causing lattice expansion. Figure 1h illustrates the interfacial crystal model, revealing a 1.8% lattice mismatch between quintuple octahedral spacings (dδ2 = 15.82 Å) of the δ-phase and quadruple octahedral spacings (dα2 = 15.54 Å) of the α-phase. Figure 1i tracks the dα2 and dδ2 evolution during aging, dδ2 increases significantly from 15.43 to 15.82 Å, corresponding to a compressive strain release process of the δ-phase as illustrated in Fig. 1c. The rise in dδ2 rapidly increases α/δ lattice mismatch to 2.14%, augmenting tensile strain in the α-phase, demonstrating that the tensile strain in the perovskite primarily originates from α/δ interfacial interactions. Initially, dα2 and dδ2 values are nearly identical due to the δ-phase precipitating from and inheriting the parent α-phase’s crystal characteristics, confirming the in-plane interface functions as the transitional initiation site for perovskite phase transformation. Consequently, δ-phase [001] orientation drives widespread formation of unfavorable in-plane interfaces between δ-phase (110) and α-phase (011) planes; these intrinsically active phase-transformation interfaces accumulate strain that dominates perovskite degradation kinetics. Phase transition interface regulation Given the link between the δ-phase orientation and α/δ interfacial behavior, suppressing the formation of δ-phase oriented crystal planes can control the density of α/δ phase transformation interface, thereby blocking the α-phase degradation pathway. Therefore, we incorporated 2D structures into the FACs perovskite to suppress phase transformation interfaces by controlling δ-phase orientation. Supplementary Fig. 6 illustrates the structures of conventional 2D perovskites (e. g. those based on Butylammonium (BA) and Phenethylammonium (PEA)) and 2D perovskitoid materials (such as the 2D structure based on crown ethers31,32). In contrast to conventional 2D perovskites, which feature corner-sharing octahedra, 2D perovskitoids are composed of face-sharing octahedra, structurally resembling the δ-phase of perovskite. Consequently, 2D perovskitoid materials allow for better control over the orientation of the δ-phase. Since the experiments in this work are based on the DMF/DMSO solvent system and 10% MACl additive, the δ-to-α phase transition of perovskite mainly occurs during the annealing process (see Supplementary Video 1), and unannealed perovskite films are predominantly dominated by the δ-phase, which is consistent with previous work33–35. Accordingly, we can perform XRD measurements on unannealed samples to investigate the influence of different 2D materials on δ-phase orientation (Supplementary Fig. 7). XRD results revealed that conventional 2D perovskites exert weak δ-phase orientation regulation, whereas Dibenzo-18-Crown-6 (DB18C6) derived 2D perovskitoid uniquely attenuated the diffraction intensity of δ-phase (100) plane while enhancing the δ-phase (101), (002), (102), and (201) planes, demonstrating a unique orientation regulation capability on δ-phase. Additionally, we introduced a series of crown ethers into the perovskite, including 12-crown-4 (12C4), 15-crown-5 (15C5), 18-crown-6 (18C6), dibenzo-15-crown-5 (DB15C5), dibenzo-21-crown-7 (DB21C7), and dibenzo-24-crown-8 (DB24C8), to assess the general applicability of their role in modulating the δ-phase orientation of 2D perovskitoids. The chemical structures of all crown ethers are depicted in Supplementary Fig. 8a. XRD patterns of perovskite films annealed at different temperatures are shown in Supplementary Fig. 8b-h. Notably, stable diffraction peaks corresponding to 2D perovskitoids at low angles were observed only in perovskite films doped with dibenzo-type crown ethers (DB15C5, DB18C6, DB21C7, and DB24C8). Furthermore, the orientation of the δ-phase has undergone obvious changes, with the (100) diffraction peak weakens and other peaks enhances, confirming the universality of the 2D-phase in regulating the orientation of the δ-phase. Notably, the perovskite film based on DB18C6 exhibited the most stable 2D structure (remaining intact at 200 °C) and the highest α-phase crystallization quality with (001) plane orientation. Consequently, we employed the DB18C6-based 2D perovskitoid to modulate the δ-phase orientation of FACs perovskite. In contrast, for perovskites doped with 12C4, 15C5, and 18C6, the low-angle diffraction peaks disappeared when annealed above 150 °C, suggesting the formation of an intermediate phase rather than a stable 2D perovskitoid. To further confirm the influence of 2D-phase formation on δ-phase and investigate the regulation on α/δ coherent interface, grazing-incidence wide-angle X-ray scattering (GIWAXS) analysis was performed on the low temperature (30 °C) annealed δ-phase-dominant samples. As shown in Fig. 2a, the control sample exhibits diffraction rings at q = 0.8, 1.2, 1.5, and 1.9 Å−1, corresponding to the (100), (101), (110), and (201) planes of δ-phase, respectively. The diffraction rings of the (100) and (110) planes display an arcuate distribution, indicating a pronounced preferential orientation. Azimuthal integrating (Supplementary Fig. 9a) confirms δ-phase (100) plane adopts an out-of-plane orientation, exhibiting stronger diffraction at 0° and 60° azimuthal angles, corresponding to the hexagonal structure of δ-phase. The orientation of the δ-phase (110) plane deviates from the out-of-plane direction by ~30°, matching the interplanar angle between the (100) and (110) planes and collectively verifying the preferential orientation of the δ-phase along the [001] orientation. In the α-phase, the (011) diffraction ring is observed at q = 1.4 Å−1, where an arc peak appears at a 30° azimuthal angle. This angular position aligns with that of the δ-phase (110) plane. The alignment of the α-phase (011) plane with the δ-phase (110) plane provides strong evidence supporting the formation of a coherent α(011)/δ(110) interface. Figure 2b displays the GIWAXS results of the 1% 2D-incorporated perovskite annealed at 30 °C. Two special diffraction rings emerge in the low-q region (q = 0.37, 0.74 Å−1). Using the formula q = 4πsinθ/λ, these q values correspond to 2θ values of 5.2° and 10.4°, respectively matching the (001) and (002) planes of the 2D-phase. The theoretical XRD pattern of the 2D-phase is illustrated in Supplementary Fig. 10. The characteristic (100) diffraction rings of the δ-phase are obviously weakened, and two newly emerged diffraction rings are observed at q = 1.6 and 1.8 Å−1, belonging to the (002) and (102) planes of δ-phase. The (100) and (110) planes of the δ-phase exhibit isotropic and uniform diffraction rings, indicating the significant role of the 2D-phase on regulating the crystal orientation of the δ-phase. Crystal phase orientation regulation. a, b GIWAXS patterns of FACs perovskite (a) and 1% 2D-incorporated perovskite (b) after annealing at 30 °C. c, d GIWAXS patterns of FACs perovskite (c) and 1% 2D-incorporated perovskite (d) after annealing at 150 °C. e, f XRD patterns of FACs perovskite (e) and 1% 2D-incorporated perovskite (f) at annealing temperatures of 30, 80, and 150 °C. The regulation effect of the 2D-phase on the α/δ coherent interface is further confirmed by its effect on α-phase orientation. As seen in Fig. 2a, the diffraction ring at q = 1.0 Å−1, corresponding to (001) plane of α-phase, exhibits enhanced diffraction intensity at 35° and 65° arc. This suggests that the α-phase (001) plane is oriented at angles of 35° and 65°, which is more clearly observed in the azimuthal angle integration results in Supplementary Fig. 9b. Given that the angle between the (001) and (112) planes in cubic structure is 35°, and that angle between the (001) and (211) planes is 65°, it can be inferred that the α-phase (112) plane is out-of-plane oriented for the sample annealed at low temperatures. It is noted that the (112) and (211) planes are equivalent in cubic structure. To confirm this speculation, the simulation of GIWAXS pattern for the α-phase (112) plane was conducted (Supplementary Fig. 11a). The specific simulation method is detailed in the supplementary information. The simulated pattern aligns well the experimental GIWAXS results in Supplementary Fig. 11b. Therefore, we conclude that the α-phase perovskite exhibits an (112) out-of-plane orientation, while the δ-phase shows an (100) out-of-plane orientation, which facilitates the formation of the coherent α(011)/δ(110) interface at an azimuthal angle of 30°. The configuration between (110) and (100) plane of δ-phase, and (011) and (112) plane of α-phase is illustrated in Supplementary Fig. 12. When the 2D perovskitoid is incorporated, the α-phase (001) plane displays uniform diffraction rings (Fig. 2b), indicating suppression of the (112) out-of-plane orientation of α-phase. This is attributed to the 2D-phase regulating the δ-phase prior to the δ-to-α transformation. Figure 2c and Fig. 2d display the GIWAXS patterns of the control and 1.0% 2D-incorporated films annealed at 150 °C. In the control FACs perovskite, α-phase (001) diffraction ring shows lower intensity at a 0° polar angle, indicating that the (112) plane retains some degree of out-of-plane orientation. This is consistent with the XRD results in Fig. 2e, where the (112) peak maintains high diffraction intensity even after annealing at 150 °C. In contrast, for the 2D-incorporated sample, the intensity of the α-phase (001) diffraction rings is essentially uniform, indicating significant suppression of the (112) plane orientation. This coincides with the XRD data in Fig. 2f, which shows notably weakened diffraction intensity of (112) plane even at 30 °C. The in-situ GIWAXS was used to observe the δ-to-α phase transition process and to investigate the regulation of the 2D-phase on the α/δ phase interface. The temperature-variable in-situ data of FACs and 1.0% 2D-incorporated perovskite are shown in Supplementary Fig. 13a and 13b, and the GIWAXS patterns of the perovskite annealed at temperature of 30, 80, 150, and 200 °C are presented in Supplementary Fig. 14. Obvious δ-phase (100) plane diffraction at q = 0.8 Å−1 and α-phase (001) plane diffraction at q = 0.975 Å−1 appear. Complete δ-to-α transition process is clearly observed both in Supplementary Fig. 13a and 13b. Further extraction of the azimuthal information for the α-phase (001) plane and δ-phase (100) plane from the GIWAXS results at various temperature points allowed us to plot the in-situ variation of the diffraction intensity at each azimuthal angle with temperature, and to investigate the orientation changes during the δ-to-α transition process. The in-situ azimuthal angle dependent diffraction intensity for α and δ phases FACs perovskite is presented in Supplementary Fig. 15a and 15b, showing that the α-phase (001) plane exhibits preferred orientations at angles of 35° and 65°, with the diffraction intensity gradually increasing during the δ-to-α transition. Meanwhile, the δ-phase (100) plane shows obvious out-of-plane orientation, and the diffraction intensity gradually decreases with temperature increasing. The observed orientation of the α and δ phase during the δ-to-α transition process indicates that the δ-to-α phase transition occurs via a transformation pathway from the out-of-plane orientation δ-phase (100) plane to the out-of-plane orientation α-phase (112) plane. Combined with the crystal plane configurations of α and δ phases (Supplementary Fig. 12), it confirms the existence of the α(011)/δ(110) phase transition interface as observed in the HRTEM image (Fig. 1g). The in-situ azimuthal angle dependent diffraction intensity of the 1.0% 2D-incorporated α and δ phases perovskite with temperature is shown in Supplementary Fig. 15c and 15d. The diffraction of the (001) plane of the α-phase is uniformly distributed and does not show any specific orientation, which confirms the suppressive effect of the 2D perovskitoid on the out-of-plane orientation of α-phase (112) plane. Meanwhile, the diffraction of the δ-phase (100) plane is shifted to high angle between 30 and 75°, confirming that the 2D perovskitoid effectively regulates the orientation of δ-phase (100) plane. To elucidate the regulation mechanism of 2D perovskitoid on the α/δ phase transition interface, δ-phase films with varied 2D perovskitoid contents were prepared to investigate its impact on δ-phase structural characteristics. XRD analysis (Fig. 3a) confirms all unannealed samples predominantly comprise δ-phase (≈ 90%), yet the preferred orientation of δ-phase shifts with the introduction of the 2D-phase. Consistent with the GIWAXS results shown in Fig. 2a and Fig. 2b, the control sample exhibits a strong (100) plane preferred orientation for the δ-phase, while the 2D-incorporated sample shows emergent peaks at 16.3°, 22.5°, 25.5°, and 26.4°, corresponding to the (101), (002), (102), and (201) planes, respectively (crystallographic schematics see Supplementary Fig. 16). Notably, these newly emerged planes deviate from [001] orientation alignment, demonstrating the 2D-phase disrupts the orientation characteristics of the δ-phase. Quantitative analysis of the integrated diffraction peaks intensities of δ-phase (Fig. 3b) reveals that the relative content of (100) plane for δ-phase attenuates significantly with increasing 2D-phase concentration, while the relative contents of non-[001] oriented δ-phase planes (e.g., (101), (002)) enhance progressively. The result confirms that the 2D perovskitoid can effectively inhibit the formation of the [001] oriented δ-phase planes, thereby hindering the generation of the α/δ phase transition interface. Regulation mechanism of 2D perovskitoid on α/δ interface. a XRD patterns of unannealed perovskite films with 0.5%, 1.0%, 1.5%, 2.0% and 5.0% 2D perovskitoid. b δ-phase (100), (101), (002), (102) and (201) planes diffraction peaks proportion of unannealed perovskite films with different 2D contents. c HRTEM image of the 1.0% 2D-incorporated perovskite. The scale bar in the HRTEM image is 2 nm. d–f Microstrain (d), octahedral spacing (e) and lattice energy (f) of perovskites with different 2D contents obtained by XRD refinement. g Diagram of the mechanism by which 2D perovskitoid improve the stability of the α-phase. Subsequent studies investigated the interaction between the 2D-phase and δ-phase to reveal the mechanism by which the 2D-phase regulates the orientation of δ-phase. HRTEM characterization of the 1.0% 2D-incorporated perovskite (as shown in Fig. 3c, the corresponding FFT analysis is presented in Supplementary Fig. 17) reveals that the (110) plane (d = 4.3 Å) of δ-phase and (002) plane (d = 8.1 Å) of 2D-phase form two distinct interfacial configurations: in the in-plane interface region (red dashed line), the δ-phase and the 2D-phase exhibits a 1:1 periodic matching, which can be confirmed by the coplanar structure of the δ-phase and the 2D-phase shown in Supplementary Fig. 4 and 6b. Meanwhile, in the out-of-plane interface (blue dashed line), the δ-phase and the 2D-phase exhibit a 2:1 periodic matching. This matched spacing satisfies the structural requirements for a strong δ-phase/2D-phase interaction, further corroborated by HRTEM results in Supplementary Fig. 18a and 18b. Additionally, the observed appearance of δ-phase planes with smaller lattice spacing confirms the orientation regulating effect of 2D-phase on the δ-phase (Supplementary Fig. 18c and 18d). The 2D/δ phase interface structure was further analyzed through XRD refinement, with fitted patterns shown in Supplementary Fig. 19 and microstructural parameter statistics provided in Supplementary Table 2. The calculated microstrain is illustrated in Fig. 3d. The tensile strain in the α-phase decreases with increasing concentration of 2D-phase, indicating enhanced α-phase stability. In contrast, the lattice strain of the δ-phase shifts dramatically from an initial compressive strain (−1%) to a significant tensile strain (+8%), demonstrating a profound impact of the 2D-phase on the δ-phase. Further analysis of in-plane interface structural parameters (Fig. 3e) reveals that the octahedral spacing (d2D = 16.5 Å) of the 2D-phase is obviously larger than that of the δ-phase (dδ2 = 15.83 Å), resulting in a 4.24% lattice mismatch. The mismatch is the origin of the substantial tensile strain experienced by the δ-phase at the interface. As the 2D-phase concentration increases, the dδ2 of δ-phase gradually increases from 15.81 to 15.90 Å, aligning with the evolving trend of lattice strain. while the d2D of 2D-phase concurrently decreases from 16.50 to 16.44 Å, gradually approaching the dδ2. The result indicates that the strain at the δ/2D-phase interface is continuously enhanced with increasing 2D-phase content, increasing interfacial strain drives the structural parameters of both phases closer. Since the δ-phase at the in-plane interface is [001] oriented, the accumulated interfacial strain impedes the formation of [001] oriented δ-phase planes, thereby promoting the growth of δ-phase with other alternative orientations. In contrast, the >5% lattice mismatch between the 2D-phase and α-phase results in minimal structural influence on the α-phase, thus, the enhanced the α-phase stability primarily stems from suppressed formation of phase transition interfaces. To more accurately assess the stability of α-phase, lattice energy calculations based on the Phillips-van Vechten-Levine (P-V-L) theory were conducted to provide further thermodynamic insight. (As shown in Fig. 3f, the method is described in the supplementary information, and the lattice energy results are summarized in Supplementary Table 3). In the control sample, the lattice energy of α-phase (3,427 kJ mol−1) is lower than that of δ-phase (3,444 kJ mol−1), consistent with α-phase metastability. However, introducing the 2D-phase reduces the lattice energy of δ-phase to 3,351 kJ mol−1, significantly below that of α-phase. This energy barrier inversion renders the α-phase the thermodynamically stable. Building upon aforementioned findings, Fig. 3g illustrates the mechanism by which 2D perovskitoid enhance α-phase stability. In the pristine perovskite system, the in-plane interface between the α-phase (011) plane and the δ-phase (110) plane initiates phase transition, where interfacial tensile strain from lattice mismatch triggers preferential α-phase transformation. Upon introducing 2D perovskitoid, a 2D/δ phase in-plane interface forms between the 2D-phase (002) plane and δ-phase (110) plane. The substantial epitaxial strain arising from the 4.24% lattice mismatch at the interface disrupts the preferred [001] orientation of δ-phase, effectively suppressing formation of the α/δ phase-transition interface. This targeted interface regulation ultimately blocks the α-to-δ transformation pathway, thereby significantly enhancing the α-phase perovskite stability. Photoelectric performance and stability To validate the stabilizing effect of 2D perovskitoid on FACs perovskite, aging studies were conducted on perovskite films with varying 2D-phase content. As shown in XRD patterns (Fig. 4a), the control sample only exhibits the α-phase, while distinct (001) and (002) diffraction peaks were observed at 5.3° and 10.6° in 2D-incorporated samples, respectively, with intensity scaling up with 2D-phase concentration. The positive effect of the 2D-phase on the α-phase stability was verified through a 1000 h humidity testing (70% RH). As shown in Fig. 4b, the control sample underwent near-complete transformation to δ-phase (≈ 98% content), while the 0.5% 2D-incorporated sample retained predominantly α-phase structure (δ-phase ≈7%). Higher 2D concentrations even completely suppressed δ-phase formation. The impact of 2D-phase on the lattice strain of α-phase was quantitatively analyzed using the Williamson-Hall (W-H) method36. As illustrated in Fig. 4c, the 2D-phase incorporation significantly reduced the α-phase microstrain by 80% (from 0.125% in control sample to 0.024% in the 5% 2D-incorporated sample), confirming its role in lattice relaxation and structural stabilization. The visual evolution of the perovskite during high humidity aging (70% RH) is depicted in Fig. 4d, the control sample transitioned to yellow δ-phase within 400 h, whereas 1.0% 2D-incorporated perovskite maintained α-phase stability after 1,000 h, confirming the enhanced stability in the 2D-incorporated perovskite. Additionally, 1% 2D-incorporated perovskite, annealed at a low temperature of 60 °C to produce certain amounts of δ-phase intentionally, was constructed for accelerated aging (Fig. 4e). After 1,000 h aging, the δ-phase content increased only from the initial 66% to 73%, with the phase transformation rate significantly lower than that of the control sample (Fig. 1a), further evidencing suppressed interfacial phase transition in the 2D-incorporated perovskite. Stability of perovskite films. a XRD patterns of annealed perovskite films with 0.5%, 1.0%, 1.5%, 2.0% and 5.0% 2D perovskitoid. b XRD patterns of perovskite films after aging at 70% RH for 1,000 h. c Microstrain calculated by W-H method of perovskite films with different 2D content. d Photographs of FACs and 1% 2D-incorporated perovskite films during aging. Sample size: 20 × 20 mm. e XRD patterns of FACs perovskite film with 1% 2D perovskite annealed at 60 °C for 5 min (blue line) and the identical film post 1000 h aging at 70% relative humidity (red line). Inset: Photographs of 20 × 20 mm perovskite film (corresponding conditions). In addition, 2D perovskitoid provides effective defect passivation in FACs perovskites, as evidenced by X-ray photoelectron spectroscopy (XPS) analysis (Supplementary Fig. 20). The 2D-phase incorporation reduces Pb 4 f binding energy, which is originates from the negatively charged cavity structure of the DB18C6 molecules in the 2D perovskitoid. DB18C6 molecules could anchor the uncoordinated Pb2+ defects and increase electron cloud density around Pb atoms. Complementary photoluminescence (PL) and time-resolved photoluminescence (TRPL) spectra confirm the defect passivation effect of 2D-phase (Supplementary Fig. 21). Incorporating <1.5% 2D-phase markedly enhances the PL intensity and carrier lifetime of the perovskite, indicating reduced defect density. However, ≥1.5% 2D-phase incorporation introduces detrimental deep-level trap states from excess insulating phase, accelerating carrier recombination. Scanning electron microscopy (SEM) characterization (Supplementary Fig. 22) further demonstrates that the optimal 0.5% 2D-phase concentration optimizes perovskite grain growth and improved morphological uniformity, effectively mitigating non-radiative recombination at grain boundaries and contributing to improving efficiency of the PSCs. To systematically assess the modulating effect of 2D perovskitoid on the photovoltaic performance of FACs-based PSCs, devices with FTO/ SnO2/ perovskite (FACs+2D)/ Spiro-OMeTAD/ Au architecture were fabricated (Fig. 5a). Current density-voltage (J-V) characteristic curve that includes both forward and reverse scans (Fig. 5b) demonstrate significantly enhanced performance for the 2D-incorporated device, achieving a PCE increase from 23.62% for control device to 25.61% for 1.0% 2D-incorporated device (third-party certification is 25.49%, as shown in Supplementary Fig. 23). The hysteresis index (HI) was calculated using the formula: HI = (PCEreverse - PCEforward) / PCEreverse. The calculated HI for the 2D-incorporated PSCs is HI2D = 0.0375, which is lower than that of the control device (HIcontrol = 0.0461). Corresponding external quantum efficiency (EQE) measurements (Fig. 5c) show improved integrated JSC from 25.05 mA cm−2 for control device to 25.32 mA cm−2 for 1.0% 2D-incorporated device, consistent with the J-V results. Steady-state power output (SPO) is shown in Fig. 5d, the control device achieved a stabilized PCE of 23.44% at 0.98 V, while the 1.0% 2D-incorporated device attained a PCE of 25.48% at 1.04 V, verifying the reliability of PCE results. Statistical analysis of 20 devices (Fig. 5e) confirms the positive effect of 2D incorporation, elevating average PCE from 23.26% for control device to 25.12% for 1.0% 2D-incorporated device. However, excessive 2D-phase (>1.5%) induces performance degradation due to insulating phase accumulation (Supplementary Fig. 24). Photovoltaic performance. a Device structure of perovskite solar cells. b Champion J-V curves of the control and the 1.0% 2D-incorporated PSCs with forward and reverse scans. c EQE spectra (left axis) and integrated JSC (right axis) of the control and the 1.0% 2D-incorporated PSCs. d Steady-state JSC and corresponding steady conversion efficiency of the control and the 1.0% 2D-incorporated PSCs. e PCE statistic plot of the control and the 1.0% 2D-incorporated PSCs. The number of statistical samples for each group is 20. For the box chart, the box is determined by the 25th and 75th percentiles coupled with median lines and mean values (small open square symbol within the box). f Operational stability at 70 ± 5% RH atmosphere of the control and the 1.0% 2D-incorporated PSCs. The error bar is the standard deviation of Four devices. Data are presented as mean values ± standard deviation. g PCE of the encapsulated control and 1.0% 2D-incorporated devices measured at maximum power point under continuous one-sun illumination at 25 °C in ambient air (35 ± 5% RH). Device stability testing confirms the enhancement effect of the 2D-phase on perovskite stability. Figure 5f shows the long-term operational stability of unencapsulated perovskite devices under ambient conditions with 70% RH. The control device degraded to 50% initial PCE after 200 h (T50 < 200 h), while the 1.0% 2D-incorporated device retained 90% initial PCE even after 1,000 h (T90 > 1,000 h). Although higher 2D-phase content further improves device stability (Supplementary Fig. 25), optimal efficiency-stability balance requires maintaining the concentration of 2D-phase ≤ 1.0%. Figure 5g illustrates the operational stability of the devices at the maximum power point tracking (MPPT). The MPPT tests were performed under encapsulated conditions in a standard white LED lighting simulating sunlight at 25 °C. The 1.0% 2D-incorporated device maintained 89.8% initial efficiency through 1000 h of continuous MPPT, demonstrating exceptional operational stability. To verify the broad applicability of the 2D perovskitoid strategy in suppressing the α-to-δ phase transition in perovskite, we implemented this approach to inverted PSCs with the structure ITO/ NiOX/ SAMs/ Cs0.05FA0.8MA0.15PbI2.88Br0.12 (CsFAMA)/ C60/ BCP/ Ag. As presented in Supplementary Fig. 26a and 26b, 1% 2D-incorporated device achieves a PCE of 23.47%, compared to 22.57% for the control device, while the average PCE increases from 21.22% to 21.68%. These performance improvements confirm the general effectiveness of the 2D perovskitoid strategy in enhancing PSCs photoelectric performance. Furthermore, XRD patterns of CsFAMA perovskite films before and after aging at 70% RH are shown in Supplementary Fig. 26c and 26d. Prior to aging, the α-phase CsFAMA perovskite exhibits a strong (112) out-of-plane orientation, while the introduction of the 2D-phase alters the structure of the α-phase and suppresses its (112) plane orientation, which is consistent with the observations in FACs perovskites (Fig. 2c,f). After 10 days of aging under 70% RH, the control CsFAMA perovskite has partially aged into the δ-phase, as evidenced by the emergence of the δ-phase (100) diffraction peak. In contrast, the 1% 2D-incorporated CsFAMA perovskite shows no evidence of phase transition and remains stable in the α-phase. These results confirm that the strategy of disrupting the α/δ phase interface through 2D perovskitoid incorporation provides a generally applicable approach for inhibiting the α-to-δ phase transition in perovskite materials. Discussion In this work, XRD refinement and HRTEM characterization were employed to reveal that the orientation of δ-phase (110) plane promotes the formation of detrimental in-plane coherent interfaces with α-phase (011) plane. Lattice mismatch at this α/δ coherent interface generates tensile strain, rendering it as an active site for α-to-δ phase transition and α-phase degradation. Targeting this phase transition interface, we introduced 2D perovskitoid into the precursor solution, leveraging structural matching between 2D-phase (002) and δ-phase (110) planes to construct a 2D/δ phase interface. Microstructural analysis demonstrates that 4.24% lattice mismatch at this 2D/δ phase interface generates sufficient epitaxial stress to disrupt the orientation of δ-phase (110) plane, thereby effectively suppressing the formation of α/δ phase transition interface and significantly enhancing α-phase stability. The FACs-based PSCs implementing this strategy achieved a PCE of 25.61%, with unencapsulated devices retaining 90% initial efficiency after 1,000 h aging at 70% RH. This interface regulation strategy fundamentally resolves the α-phase instability of FA-based perovskites while providing a universal guidance for improving the α-phase stability of other perovskite systems. Methods Materials Formamidinium iodide (FAI, 99.99%) and Methylamine hydrochloride (MACl, 98%) were purchased from Xi’an Polymer Light Technology. Lead iodide (PbI2, 99.999%), Cesium iodide (CsI, 99.999%), 2,2’,7,7’-Tetrakis [N,N-di (4-methoxyphenyl)amino]−9,9’-spirobifluorene (Spiro-OMeTAD, 99.8%) were purchased from Advanced Election Technology Co., Ltd. N, N-dimethylformamide (DMF, 99.8%), 4-tertbutylpyridine (4-TBP, 99.8%), dimethyl sulfoxide (DMSO, 99.9%), lithium bis(trifluoromethanesulfonyl)imide (Li-TFSI, 99.95%), chlorobenzene (CB, 99.8%) and crown ethers (12C4, 15C5, 18C6, DB15C5, DB18C6, DB21C7 and DB24C8) were purchased from Sigma-Aldrich. The SnO2 colloid solution (tin (IV) oxide, 15% in H2O colloidal dispersion) was purchased from Alfa Aesar. Solar cell substrates are pre-patterned FTO glass obtained from Advanced Election Technology Co., Ltd. Fabrication of perovskite solar cells Perovskite solar cells were fabricated on pre-patterned FTO glass substrates. The substrates were cleaned by deionized water, acetone, and isopropanol and ethanol, each step for 30 minutes. Prior to use, the FTO coated glass was dried using a nitrogen gun and treated with UV ozone for 20 minutes to improve wettability. Subsequently, SnO2 colloidal solution (15%) was diluted with deionized water (1:3, v/v) and spin-coated at 4,000 rpm for 30 s then annealed at 150 °C for 30 min. To prepare the FA0.9Cs0.1PbI3 active layer, a 1.45 M FA0.9Cs0.1PbI3 precursor solution was prepared by dissolving stoichiometric FAI, CsI, PbI2, MACl with 0.9: 0.1: 1.05: 0.10 molar ratio in DMF:DMSO (4:1). For the preparation of perovskite containing 2D perovskitoid, it is sufficient to add DB18C6 as an additive (at concentrations of 0.5%, 1%, 1.5%, 2% and 5%) to the configured perovskite precursor solution. The perovskite precursor solutions were spin-coated on the SnO2 film at 1000 and 4000 rpm for 10 and 20 s, respectively. 0.3 mL of ethyl acetate was slowly dripped on the rotating substrates about 5 seconds before the end of the spin-coating. The black thin films were formed after annealing in a hot plate at 150 °C for 10 min. Then, the Spiro-OMeTAD solution, which was prepared by dissolving 72.3 mg Spiro-OMeTAD in 1 mL chlorobenzene, 28.8 μL 4-tertbutylpyridine solution, and 17.5 μL Li-TFSI solution (520 mg mL−1 in acetonitrile), was deposited by spin-coating at 4000 rpm for 30 s. The films were left in an oxygen atmosphere with a relative humidity of about 15% for 12 h to oxidize Spiro-OMeTAD. Lastly, 80 nm layer of Au was thermally evaporated under vacuum, and the device size areas were 0.09 cm2. Photovoltaic performance characterization The performance of perovskite solar cells was measured using a solar simulator (SAN-EIELECTRIC XES-40S2-CE). The light intensity was calibrated with a reference Si cell (KG3, Newport). Under AM1.5 G 1 sun illumination in ambient air at room temperature, an external voltage bias was applied, and the current response was detected using a Keithley 2400 digital source meter, allowing for the measurement of the current-voltage (J-V) characteristics. The voltage scan rate, scan steps, and delay time were 100 mV s−1, 20 mV, and 100 ms, respectively. The active area of the devices was defined by using a black mask with an aperture area of 0.09 cm2. No light soaking or voltage bias was applied before the measurement. The EQE spectrum was obtained using a QTest Station 2000ADI system (Crowntech Inc.). The MPPT measurements were conducted under encapsulation conditions, utilizing standard white LED illumination. The light intensity of the LED lamps used in the measurement system was 100 mW cm−2, and this was verified with a reference silicon photodiode. Film characterization The GIWAXS measurements were conducted using the Xeuss 2.0 from the French company XENOCS, employing X-rays with a wavelength of λ ≈ 1.54 Å at an incident angle of 1°. The XRD patterns of the perovskite were acquired using a Bruker D8 advance diffractometer with Cu Kα (λ¼, 1.54178 Å). The PL and TRPL spectra were obtained by DeltaFlex fluorescence spectrometer (HORIBA). The XPS measurements were conducted by ESCALAB 250Xi, Thermo SCIENTIFIC. The SEM images were obtained by field emission scanning electron microscopy (FESEM, JEOL 7100 F). The HRTEM images were obtained by using Talos F200X field emission transmission electron microscope (Thermo Fisher Scientific) with probe and spherical error corrector on the perovskite sample prepared by FIB. All images have a resolution of 120 pm and are obtained at an accelerated voltage of 200 kV. The perovskite film to be measured is first cut using a dual-beam focused ion beam (FIB) device (Helios 5 UC, Thermo Fisher Scientific), using a low-voltage mode (5 kV), cutting the sample to a thickness of 50 nm. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Video 1 Reporting Summary Transparent Peer Review file Acknowledgements The authors thank the spherical aberration-corrected TEM center of Hubei University for relevant characterizations. The authors also thank the Analytical and Testing Center of the School of Materials Science and Engineering, Hubei University for their assistance with material characterization and testing. Author contributions D.F.W. and C.W.X. conceived the project; Y.Z.L. made the devices and contributed to the performance improvement; C.W.X. measured the stress of perovskites and performed the results analysis; C.W.X. and Q.H.B. conducted the GIWAXS measurements and performed the results analysis; M.H., Z.Q.L., and Y.B.Z. participated in the characterizations of devices; C.W.X., Q.H.B. and D.F.W. contributed to the results analysis. C.W.X., D.F.W. and Z.K.L. wrote and revised the manuscript; D.F.W., T.J.Z. and Z.K.L. supervised the project. All authors discussed the results and commented on the final manuscript. Peer review Peer review information Nature Communications thanks Zeguo Tang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Funding D. F. W. discloses support for the research of this work from the National Natural Science Foundation of China (grant nos. 11174071, 11304088, and 51372180). T. J. Z. discloses support for publication of this work from the National Natural Science Foundation of China (grant nos. 12374083). Z. K. L. discloses support for the research of this work from the Shaanxi Provincial Science Fund for Distinguished Young Scholars (2025-JC-JCQN-004). Data availability The data supporting the findings of this study are available within the published article and its Supplementary Information. Source data are provided with this paper and at 10.6084/m9.figshare.30484886 (ref. 37). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Zhike Liu, Email: zhike2015@snnu.edu.cn. Tianjin Zhang, Email: zhangtj@hubu.edu.cn. Duofa Wang, Email: wangdf@hubu.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73739-0. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Video 1 Reporting Summary Transparent Peer Review file Data Availability Statement The data supporting the findings of this study are available within the published article and its Supplementary Information. Source data are provided with this paper and at 10.6084/m9.figshare.30484886 (ref. 37).