Introduction

Formamidinium lead iodide (FAPbI3) is a leading candidate for high-efficiency solar cells due to its near-ideal bandgap and superior stability1,2,3,4. However, its practical application is fundamentally limited by the challenge of its crystallization. The required transition from a non-photoactive δ-phase to the desired photoactive α-phase is kinetically unstable and often incomplete5,6,7. This results in films plagued by residual δ-phase impurities and high defect densities, which ultimately compromises both device performance and operational stability, making the precise control of the δ-to-α phase transformation the critical challenge8,9,10.

Numerous efforts have been dedicated to controlling the crystallization kinetics of FAPbI3, aiming to achieve the desired black α-phase and enhance device performance11,12,13. These strategies include manipulating precursor adducts14,15, optimizing solvent16,17,18 and anti-solvent systems19,20, controlling intermediate phase formation5,21,22,23,24, and employing various deposition techniques25,26. However, a common limitation of these strategies, especially in the widely used one-step method coupled with anti-solvent treatment, is the difficulty in achieving precise control over the crystallization process27,28. While anti-solvents accelerate nucleation, they can also lead to uncontrolled crystal growth. This is further exacerbated by the inherently top-down crystallization process of perovskites, where rapid solvent evaporation from the wet film surface leads to preferential nucleation at the surface, potentially resulting in non-uniformity and a higher density of defects29. This top-down crystallization process, while posing challenges, also presents a unique opportunity for manipulating the perovskite film formation. Previous research has shown that top-surface modulation is more effective than traditional additive engineering in reconstructing the perovskite lattice and achieving precise control over crystal quality30,31,32,33. By acting as a template, the modified perovskite film surface can induce epitaxial growth and promote the formation of well-crystallized, low-defect perovskite domains34,35,36,37. Therefore, it is crucial to develop strategies that effectively leverage this top-down crystallization process to guide the formation of a highly ordered perovskite crystal structure while minimizing defects.

Self-assembled molecules (SAMs) have recently become a powerful tool in perovskite solar cell engineering, especially as hole selective layers (HSLs) in inverted (p-i-n) devices38,39,40. As HSLs in p-i-n structures, SAMs effectively reduce interfacial recombination and enhance hole extraction by modifying work function, passivating defects, and improving interface contact41,42,43. However, unlike their established success in p-i-n devices, the use of SAMs to direct perovskite crystallization in conventional (n-i-p) perovskite solar cells remains largely unexplored because inherent structural differences in the n-i-p architecture, particularly the position of the HSL above the perovskite layer, limit the realization of effective SAM anchoring and uniform coverage required for crystallization control.

To overcome this challenge, this work introduces a pioneering top-down strategy that leverages the incorporation of SAMs into the anti-solvent to direct the ordered crystallization of α-FAPbI3 in the n-i-p architecture. We demonstrate that a rationally designed SAM (Synthesis route in Supplementary Fig. 1), which consists of a carbazole core, a flexible butane linker, and a terminal anchoring group, spontaneously forms an ordered, self-assembled layer on the nascent perovskite surface during film formation. This in-situ formed self-assembled layer then acts as a dynamic template, guiding the subsequent top-down crystallization process to yield highly ordered α-phase films (Fig. 1a). Employing a series of in-situ measurements, we rigorously demonstrate that the incorporated SAMs effectively suppress undesirable intermediate phases, accelerate the efficient δ-to-α phase transition, and strategically slow down crystallization, ultimately promoting highly ordered perovskite film growth. A systematic investigation comparing the -SCN group to other anchoring groups (-NH2, -CN, -COOH) confirms that this self-assembly-driven crystallization control is unique to the -SCN variant, owing to its superior interaction with the perovskite lattice. This meticulously designed strategy culminates in high-quality, ordered perovskite films that exhibit significantly enhanced efficiency and stability, yielding a champion power conversion efficiency (PCE) of 26.18% (Certified PCE = 25.67%) for single-junction devices and 21.28% for mini-modules, thus demonstrating the significant potential of SAMs in controlling perovskite crystallization within the n-i-p architecture.

Fig. 1: Morphological and structural characterization of SAM based perovskite films.
Fig. 1: Morphological and structural characterization of SAM based perovskite films.The alternative text for this image may have been generated using AI.
Full size image

a Schematic diagram of the self-assembled molecule (SAM) directed the crystallization of the perovskite film during the fabrication process. SAM (Cz-SCN) is introduced via the anti-solvent to form a templating layer on the nascent film surface that directs ordered perovskite crystallization. b Cross-sectional scanning electron microscopy (SEM) images and c Top-view SEM images. Scale bar, 1 μm. PbI2 is marked by blue circles. d 2D Grazing-incidence wide-angle X-ray scattering (GIWAXS) patterns and corresponding e 1D out-of-plane intensity plots. The incident angle was set as 0.10°. f Time-of-flight secondary-ion mass spectrometry (ToF-SIMS) depth profiles and g corresponding ion images of the SAM-treated perovskite film. The profiles show the distribution of characteristic ions for perovskite components (NH4+, FA+, Pb+), SnO2 (Sn+), and the SAM fragment (Cz+).

Results

Self-assembled molecular templates for ordered perovskite growth

9-(4-thiocyanatobutyl)−9H-carbazole (Cz-SCN) (denoted as SAM in Figs. 1 and 2) was successfully synthesized and incorporated into the anti-solvent (optimal 10 mmol/L in Ethylacetate (EA), Supplementary Fig. 2, Supplementary Note 1) to manipulate perovskite crystallization. Cross-sectional scanning electron microscopy (SEM) images (Fig. 1b) reveal a significant difference in perovskite film morphology upon SAM treatment (w/SAM). The irregular crystals observed in the control film are replaced by monolithic grains spanning the entire film thickness, suggesting enhanced crystal growth and a reduction in grain boundaries. Top-view SEM images (Fig. 1c) and grain size distribution histograms (Supplementary Fig. 3) further confirm this enhancement, showing an increase in average grain size from 0.5 to 1.3 μm and a complete suppression of residual PbI2 impurities in the w/SAM film. Furthermore, atomic force microscopy (AFM) images reveal a significant decrease in surface roughness for the w/SAM perovskite films, from 23.9 to 18.8 nm (Supplementary Fig. 4). Collectively, this transition to larger, phase-pure grains with fewer boundaries and a smoother surface creates a highly ordered morphology. This is crucial for superior device performance as it minimizes defect-driven recombination losses while providing unobstructed pathways for efficient charge transport and extraction44,45. Grazing-incidence wide-angle X-ray scattering (GIWAXS) measurements performed at different incidence angles offer further insights into how SAM affects perovskite crystallization (Fig. 1d and Supplementary Fig. 5)46. The w/SAM perovskite film exhibits a pure black α-phase with no detectable PbI2 signals, compared to the control counterpart. Notably, the w/SAM film displays a distinct signal at qz = 0.13 Å⁻¹ along the direction perpendicular to the substrate, corresponding to a lamellar spacing of approximately 4.5 nm (Fig. 1e and Supplementary Fig. 6). This observation suggests the presence of ordered lamellar structures within the film, potentially arising from the self-assembly of SAM molecules. Increasing the concentration of the SAM reveals additional lamellar stacking peaks at qz = 0.13,0.35, 0.57, and 0.79 Å⁻¹ (Supplementary Fig. 7)47. This indicates a preferential edge-on orientation of SAMs relative to the substrate, facilitating efficient vertical charge transport across the perovskite film. To further investigate the spatial distribution of the SAM within the perovskite films, we employed time-of-flight secondary-ion mass spectrometry (ToF-SIMS) (Fig. 1f, g). The SIMS depth profiles, acquired in positive ion mode, revealed characteristic fingerprint ions, including NH4+ (m/z = 18), FA+ (m/z = 45), GA+ (m/z = 69), Sn+ (m/z = 120), Pb+ (m/z = 208), and a fragment of SAM (Cz+, m/z = 225, corresponding to the carbazole moiety after loss of SCN). These fingerprint ions provide direct evidence for the presence of SAM and perovskite components throughout the film. Furthermore, the cross-sectional ToF-SIMS images confirmed a uniform distribution of Cz-SCN throughout both the bulk and surface of the perovskite film22. This homogeneous distribution suggests that SAM effectively permeates the perovskite layer during film formation, potentially influencing crystallization throughout the entire film. It should be noted, however, that the apparent uniformity in ToF-SIMS depth profiles mainly reflects fragment distribution generated during sputtering, together with depth resolution limits that can smear sharp interfacial features and broaden organic signals. Consistent with this interpretation, UPS measurements show a clear increase in work function at the Cz‑SCN‑modified top surface but negligible change at the buried perovskite interface, indicating minimal SAM accumulation there (Supplementary Fig. 8). In addition, SEM analysis of the buried interface reveals improved crystallinity, which is attributed to more complete perovskite crystallization induced by SAM (Supplementary Fig. 9). Ultraviolet–visible (UV–vis) absorption spectra indicate that SAM treatment does not alter the bandgap of the perovskite films (Supplementary Fig. 10).

Fig. 2: In-situ analysis of the SAM-directed perovskite crystallization process.
Fig. 2: In-situ analysis of the SAM-directed perovskite crystallization process.The alternative text for this image may have been generated using AI.
Full size image

a, b In-situ GIWAXS patterns of control and SAM-treated perovskite films during spin-coating and annealing. Spin-coating lasted 40 s (10 s at 1000 rpm, then 30 s at 5000 rpm), with 120 μL ethyl acetate (EA) dripped as antisolvent during the final 10 s (30–40 s). Data from this interval were selected, and annealing is shown immediately afterward, starting at 40 s. c Temporal evolution of the α/δ phase intensity ratio derived from the GIWAXS data. d Density functional theory (DFT) calculation of the surface formation energy for the α-FAPbI3 perovskite with and without the SAM. e, f In-situ photoluminescence (PL) spectra for the control and SAM-treated perovskite films and g, h corresponding magnified view of the 40–100 s annealing period. i, j In-situ Ultraviolet–visible (UV–vis) absorption spectra for the control and SAM-treated perovskite films and k, l corresponding magnified view of the 40–100 s annealing period.

In-situ probing of the SAM-directed crystallization pathway

Figure 2 provides a comprehensive analysis of the perovskite crystallization process using in-situ GIWAXS, in-situ photoluminescence (PL) spectroscopy, and in-situ UV–vis absorption spectroscopy, revealing the profound impact of the SAM on nucleation, crystal growth, and phase transformation. In-situ GIWAXS measurements (Fig. 2a, b and Supplementary Note 2) reveal distinct crystallization pathways for the control and w/SAM perovskite films. During spin-coating, after the anti-solvent dripping step, the control film exhibits signals corresponding to the hexagonal 2H (δ-phase, qz = 0.80 Å−1), hexagonal 4H polycrystal (qz = 0.84 Å−1), and 3 C (cubic) phase (qz = 1.0 Å−1), indicating a heterogeneous mixture of intermediate phases48,49. In contrast, the w/SAM perovskite film predominantly shows the 2H phase, suggesting that SAM influences the initial perovskite nucleation, promoting a more uniform intermediate phase. During annealing, the in-situ GIWAXS data reveals two distinct stages: Stage I, representing the δ- to α-phase transition as residual solvents evaporate, and Stage II, corresponding to the growth and stabilization of the α-phase perovskite. In the control film, the δ- to α-phase transition is incomplete, with residual hexagonal 6H polytype (qz = 0.87 Å⁻¹) persisting even after 260 s of annealing (Fig. 2a). The presence of MA2Pb3I8·2DMSO during Stage I further complicates the crystallization process and hinders the complete phase transformation, as our previous report22. In contrast, the w/SAM perovskite film exhibits complete conversion to the α-phase within 35 s (Fig. 2b). Intriguingly, a strong signal appears at qz = 0.13 Å⁻¹ during Stage II, indicative of ordered lamellar structures formed by SAMs. During the early stage of annealing, rapid solvent loss and preliminary crystallization produce a disordered grain structure. Continued thermal input drives the SAM toward an ordered arrangement that guides crystal growth. This spatiotemporal evolution, driven by thermodynamic stabilization and facilitated by molecular diffusion, explains why long-range lamellar stacking appears only after 20–30 s of annealing, as evidenced by Bragg diffraction spots in the GIWAXS pattern. These results indicate that SAM-assisted crystallization requires both thermal activation and structural relaxation before exerting its templating effect on perovskite grain growth. This observation confirms that SAM self-assembly occurs concurrently with the α-phase transition, suggesting a templating effect on crystal growth. This process likely disrupts the face-sharing octahedral structure (6H) of the δ-phase, facilitating the transition to the corner-sharing architecture (3C) of α-FAPbI3. The formed surface crystallization of α-FAPbI3 serves as a template, propagating the phase transition towards the bulk of the film. The time evolution of peak intensity ratios between the α-phase and δ-phase (α/δ) further highlights the impact of SAM. The steeper slope for the w/SAM film indicates a faster and higher transformation rate from δ to α-phase (Fig. 2c). Moreover, the near-linear evolution of the α/δ ratio for the w/SAM film suggests a more direct conversion pathway, whereas the staged evolution observed in the control film indicates competing transformation processes, involving both δ- to α-phase transition and the conversion of MA2Pb3I8·2DMSO to the α-phase22. Density functional theory (DFT) calculations provided further insights into the thermodynamics of perovskite crystallization and phase transition (Fig. 2d)50. The presence of SAM significantly reduced the surface energy of α-phase perovskite (from 3.32 to 1.45 eV), thermodynamically driving the preferential formation of the black phase (Supplementary Fig. 11 and Supplementary Note 3).

In-situ PL measurements reveal complex nucleation and growth dynamics of the control perovskite film (Fig. 2e and Supplementary Note 4)51,52. Upon antisolvent dripping, an intense, broad PL peak emerges at ~700 nm, attributed to Methylamine-rich (MA-rich) nanocrystal nuclei. This peak rapidly redshifts to ~735 nm within 8 s as FA-rich nuclei form and grow. The consistently large full-width at half-maximum of the emission throughout this initial stage indicates a highly polydisperse size distribution of these luminescent species. The subsequent annealing process is marked by three distinct and dramatic stages. As shown in the enlarged PL spectrum in Fig. 2g, during the initial 15 s, the PL intensity undergoes pronounced oscillations as it redshifts towards 760 nm. This heterogeneous behavior reflects a competition between PL enhancement from crystalline growth and severe quenching caused by defect generation during the disordered coalescence of nanocrystals. Second, the intensity becomes severely quenched for an extended period (~55–100 s). This prolonged PL quenching, consistent with the incomplete δ-to-α phase transformation observed in the in-situ GIWAXS data, is attributed to the presence of the photoinactive δ-phase, which quenches the α-FAPbI3 luminescence. Finally, after 60 s annealing, a gradual PL recovery occurs as grains coarsen. However, the PL intensity does not fully recover its initial peak value due to the persistent presence of defects and grain boundaries in the fully annealed film. The crystallization dynamic of the w/SAM film presents a stark contrast, indicative of a highly ordered process (Fig. 2f, h). Upon antisolvent treatment, no MA-rich emission is observed. Instead, a single, weaker PL peak appears at ~680 nm, which can be attributed to an intermediate phase involving the SAM and perovskite precursors, consistent with our in-situ UV–vis absorption data as discussed below. During annealing, the PL emission undergoes a smooth and continuous redshift while its intensity steadily intensifies, tracking the formation and growth of α-FAPbI353. This controlled evolution likely stems from a “slow-release” effect, where SAM molecules temporarily chelate with perovskite precursors, retarding their immediate reaction and facilitating a more orderly crystallization process. Crucially, the redshift in the w/SAM film initiates earlier (5 s) than in the control film (7 s), indicating that the SAM effectively lowers the formation energy barrier of the α-phase and accelerates the δ-to-α transformation (enlarged spectrum in Fig. 2h). Unlike the control, the w/SAM film exhibits only a single, shallow dip in PL intensity, which corresponds to the minimal non-radiative energy loss during the orderly phase transition. Subsequently, the PL intensity increases monotonically and surpasses its initial value, a direct consequence of the high-quality, defect-minimized crystal lattice formed under the guidance of the SAM monolayer.

In-situ UV-vis absorption spectroscopy (Fig. 2i, j) provides further insights into the crystallization process, particularly the role of the antisolvent in promoting nucleation. Immediately after antisolvent dripping during spin-coating, the control film exhibits several strong absorption peaks in the 410–440 nm range, attributable to various intermediate species present in the perovskite precursor solution. As the solvent evaporates and nucleation proceeds, multiple strong absorption peaks appear between 450 and 500 nm, due to the formation of multiple types of nuclei. In contrast, during the spin-coating process, the w/SAM perovskite film initially shows a single, strong absorption peak at 420 nm, corresponding to PbI42−• SAM intermediate phase. This suggests that SAM preferentially adsorbs onto the Pb-I framework, displacing iodide ions and promoting a more homogenous intermediate phase54. As spin-coating continues, a prominent peak appears at 460 nm in the w/SAM perovskite film, assigned to the δ-phase perovskite, consistent with the formation of a single dominant type of nucleus, as also observed in the GIWAXS data. During the initial annealing stage (Stage I, Fig. 2k, l), the absorption onset of the w/SAM perovskite film redshifts earlier (5 s after annealing) compared to the control film (7 s). This faster redshift, indicating an accelerated δ-to-α phase transition, is consistent with the in-situ PL observations and is attributed to Cz-SCN reducing the energy barrier for α-phase formation. Moreover, as highlighted in Fig. 2k, l the w/SAM perovskite film exhibits a near-linear redshift from 420 to 750 nm, suggesting an ordered and gradual δ-to-α phase transformation guided by the SAM. In contrast, the control film undergoes a multi-stage redshift, indicative of a disordered phase transformation arising from the presence of multiple nuclei species. Furthermore, the delayed emergence of the long-wavelength absorption beyond 770 nm in the w/SAM perovskite film, compared to the control, suggests that SAM retards the crystallization of α-FAPbI3 during Stage II. This slower, more controlled crystallization is conducive to the formation of high-quality perovskite films55. We speculate that Pb2+-SCN coordination is the dominant driving force for α-phase stabilization and templated crystallization, while π-π stacking among carbazole backbones cooperatively sustains ordered packing and enhances the robustness of the templating effect.

Mechanism of anchoring group-dependent crystallization control

To elucidate whether the successful crystallization control demonstrated by Cz-SCN is a general property of this molecular class or a specific function of this molecule, three analogous Cz-molecules (Cz-NH2, Cz-CN, and Cz-COOH) are investigated. A comparison of the resulting films reveals a stark contrast. GIWAXS and SEM analysis show that films treated with these alternative molecules not only suffer from residual PbI2 impurities but also completely lack the self-assembled lamellar structure unique to the Cz-SCN case (Fig.1d, Supplementary Figs. 4 and 12–14). This pivotal finding confirms that the ability to simultaneously ensure phase purity and drive self-organization is a capability exclusive to the Cz-SCN molecule. Quantitative GIWAXS analysis further demonstrates this distinction, showing that the Cz-SCN film exhibits the strongest and sharpest (100) diffraction peak together with the highest (100)/PbI2 intensity ratio, confirming its superior ability to suppress PbI2 formation and promote highly ordered crystallization (Supplementary Fig. 15). To understand the fundamental origin of this unique ability, a synergistic investigation is undertaken to dissect the underlying molecule-perovskite interactions. DFT calculations provide the initial theoretical analysis. By calculating the adsorption energy (Eads) of each molecule onto a perovskite surface with an iodine vacancy, we found that Cz-SCN possesses a remarkably strong adsorption energy of −0.77 eV, substantially more favorable than its counterparts (Fig. 3a)56. This superior energetic stabilization, rooted in the large dipole moment and highly negative electrostatic potential of the molecule, provides the robust anchoring force necessary to form an ordered template and resist impurity formation (Fig. 3b). Fourier-transform infrared spectroscopy (FTIR) further tracks changes in the chemical bonds of the anchoring groups (Fig. 3c). The S-C≡N stretching vibration in Cz-SCN undergoes a substantial blueshift of 10 cm−1 upon interaction with the perovskite, indicative of a strong coordination bond. In contrast, the other groups exhibit weaker interactions, reflected by smaller blueshifts for the C≡N (3 cm⁻¹) and C=O (5 cm⁻¹) stretches, while the N-H stretch remain effectively unchanged57. XPS analysis of the Pb 4f5/2 peak (Fig. 3d, Supplementary Fig. 16 and Supplementary Note 5) for the Cz-molecule-treated films reveal a downshift in binding energy, indicative of a Lewis acid-base interaction with undercoordinated Pb2+ ions. Cz-SCN induces the largest energy shift (0.41 eV), significantly greater than the shifts from Cz-COOH (0.30 eV), Cz-CN (0.19 eV), and Cz-NH2 (0.09 eV). To further substantiate this observation, XPS measurements of the relevant core levels (N 1s and O 1s) reveal clear binding energy shifts upon SAM introduction (Supplementary Fig. 17 and Supplementary Note 6). Among them, Cz-SCN shows the largest shift (N 1s: +0.46 eV), indicative of strong Pb²⁺-SCN coordination. The varying interaction strengths are corroborated by complementary characterizations. UV-Vis absorption spectroscopy (Supplementary Fig. 18) shows that Cz-SCN treatment most effectively suppresses the characteristic absorption of PbI2-related species, which is visually confirmed by the bleaching of the yellow PbI2 film to transparent (Supplementary Fig. 19). Consistently, XRD analysis (Supplementary Fig. 20) reveals that the PbI2 diffraction peak at 12.8° disappears in films prepared from Cz-SCN/PbI2 solutions, whereas control and other Cz-analogue samples still exhibit residual PbI2 signals. These results collectively suggest that PbI2 is not preserved as a separate phase but is rapidly consumed, most likely through the formation of a transient intermediate. Collectively, although the four Cz derivatives exhibit distinct aqueous pKa values (Supplementary Table 1, Supplementary Note 7), the direct influence of acidity/basicity on crystallization is limited in the non-aqueous DMF/DMSO precursor environment, where coordination strength with undercoordinated Pb2+ dominates.

Fig. 3: Mechanistic investigation of anchoring group-dependent crystallization control.
Fig. 3: Mechanistic investigation of anchoring group-dependent crystallization control.The alternative text for this image may have been generated using AI.
Full size image

a DFT-calculated adsorption energies (Eads) of four Cz-molecules (Cz-NH2, Cz-CN, Cz-COOH, Cz-SCN) on the perovskite surface with an iodine vacancy. b Calculated dipole moments and electrostatic potential (ESP) maps for the four Cz-molecules. c Fourier-transform infrared spectroscopy (FTIR) showing the vibrational shifts of the anchoring groups upon interaction with PbI2. d X-ray photoelectron spectroscopy (XPS) spectra of the Pb 4f5/2 peak for perovskite films treated with different Cz-molecules. e In-situ PL spectra and corresponding f PL intensity evolution at 770 nm for films treated with different SAMs during thermal annealing. In all the figures, arb. units is defined as a.u.

The profound consequences of this superior, Cz-SCN-guided crystallization are evident in the optoelectronic properties and formation kinetics of the film. The Cz-SCN-treated film exhibit the lowest trap-state density (Nt) in both hole-only or electron-only devices, the values approximately 60% lower than the control and markedly superior to the other modified films (Supplementary Figs. 21 and 22, Supplementary Tables 2–3, and Supplementary Note 8). These findings on static trap states are further corroborated by an analysis of charge carrier dynamics using time-resolved photoluminescence (TRPL) spectra (Supplementary Fig. 23 and Supplementary Note 9). The Cz-SCN-based film display a substantially longer average PL lifetime (τave) of 1980.9 ns, a 2.9-fold increase compared to the control film (677.4 ns) (Supplementary Table 4). The reduced trap density and extended carrier lifetime demonstrate that the Cz-SCN treatment effectively suppresses non-radiative recombination. To quantitatively probe the crystallization kinetics during thermal annealing, the temporal evolution of PL intensity at 770 nm was monitored in-situ (Fig. 3e, f). All as-deposited films initially exhibit a transient PL peak within the first 25 s of annealing, where the brief emission from nascent crystallites is rapidly quenched by defects as non-radiative recombination pathways. The subsequent recovery and rise in PL intensity directly track the growth and perfection of the emissive α-phase crystalline domains. The control, Cz-NH2, and Cz-CN films display a slow and inefficient PL recovery to low final intensities, indicating a crystallization pathway plagued by defect formation. While Cz-COOH offers a modest improvement, the Cz-SCN film exhibits a remarkably superior kinetic evolution. It demonstrates a rapid and sustained PL recovery to a final intensity far surpassing all other samples. This steep rise and high final efficiency provide compelling dynamic evidence that the Cz-SCN-guided process most effectively suppresses non-radiative recombination during annealing, yielding a film of the highest electronic quality. Ultimately, this well-ordered, defect-minimized interface, formed by the strong Cz-SCN dipole, also establishes the most favorable energy level alignment, upshifting the perovskite valence band maximum to an optimal −5.18 eV for barrier-free hole extraction, as confirmed by Ultraviolet photoelectron spectroscopy (UPS) (Supplementary Fig. 24)57.

Photovoltaic performance and long-term stability

N-i-p type PSCs are fabricated, employing the structure of glass/FTO/CBD-SnO2/perovskite/Spiro-OMeTAD/Au. A comparative evaluation of devices incorporating the various Cz-molecules reveal the clear superiority of the Cz-SCN treatment (Supplementary Fig. 25 and Supplementary Table 5). While control devices yield an average PCE of 23.5 ± 0.7%, the Cz-SCN-based cells achieve a significantly higher average PCE of 25.8 ± 0.4% (Supplementary Figs. 26 and 27). The champion device featuring the Cz-SCN-modified perovskite layer deliver a remarkable PCE of 26.18%, driven by an enhanced open-circuit voltage (VOC) of 1.190 V and a high fill factor (FF) of 83.65% (Fig. 4a). This performance is confirmed by a stabilized power output (SPO) of 25.9% (Supplementary Fig. 28). A device sent for external certification achieved a PCE of 25.67%, which is among the top certified efficiencies reported for n-i-p architectures (Supplementary Fig. 29 and Supplementary Table 6). The integrated short-circuit current densities (JSC) from IPCE measurements (Fig. 4b) of PSCs are within 5% mismatch of the J-V results. The enhanced VOC and FF in Cz-SCN-based devices are attributed to a synergistic improvement in device physics. Mott-Schottky and Electrochemical impedance spectroscopy (EIS) measurements collectively confirm that the Cz-SCN treatment enlarges the built-in potential, which both facilitates more efficient charge extraction and suppresses carrier recombination (Supplementary Figs. 30 and 31, Supplementary Tables 7–8, and Supplementary Notes 10–11)58,59. To demonstrate the scalability of this approach, 5 cm × 5 cm mini-modules comprising six series-connected subcells are fabricated. The champion mini-module, with an active area of 14.21 cm2, achieved a high PCE of 21.7%, with a VOC of 7.024 V, a JSC of 4.01 mA cm−2, and an FF of 77.05% (Fig. 4c). Such efficiency is lower than that of small-area devices due to inevitable series resistance and inactive area losses in module fabrication60,61.

Fig. 4: Comparison of device evaluations containing different Cz-molecules.
Fig. 4: Comparison of device evaluations containing different Cz-molecules.The alternative text for this image may have been generated using AI.
Full size image

a J-V curves of the champion devices with different Cz-molecules. b Incident photon-to-current efficiency (IPCE) spectra and corresponding integrated short-circuit current densities (JSC) for the champion devices. c Photograph and J-V curve of the champion 5 cm × 5 cm mini-module based on the Cz-SCN treatment. d Long-term stability of unencapsulated devices stored in ambient air (23 ± 3 °C, 50 ± 10% RH) (ISOS-D-1). e Photostability of unencapsulated devices under continuous 1-sun illumination in a nitrogen atmosphere (ISOS-L-1). f Operational stability of the unencapsulated control and Cz-SCN-based devices under continuous maximum power point tracking (MPPT) in a nitrogen atmosphere (ISOS-LC-1). All of the error bars in Fig. 4d, e represent the standard deviation for six devices.

The practical viability of the Cz-SCN treatment is validated through a series of rigorous long-term stability tests on unencapsulated devices. The Cz-SCN-treated PSC demonstrate exceptional environmental robustness, retaining 95% of its initial PCE after 1500 h in ambient air (23 ± 3 °C, 50 ± 10% RH). This stands in stark contrast to the control device, which catastrophically degrade to just 60% of its initial performance under the same conditions (ISOS-D-1, Fig. 4d)62. This enhanced humidity stability stems from the superior hydrophobicity of the Cz-SCN-treated surface, where the dense, uniform monolayer effectively shields the perovskite from moisture ingress (Supplementary Fig. 32). Furthermore, the Cz-SCN treatment impart outstanding photostability. Under continuous 1-sun illumination in a nitrogen atmosphere, the device maintains 90% of its initial PCE after 2000 h, outperforming all others devices (ISOS-L-1, Fig. 4e). Even more impressively, during continuous maximum power point tracking (MPPT) for 1000 h, the Cz-SCN device retain a remarkable 92% of its initial efficiency; in comparison, the control device decay to 80% of its initial value after only 450 h (ISOS-LC-1, Fig. 4f). Post-aging SEM and XRD analyses further confirm that Cz-SCN treated devices retain more compact morphology and strong α-phase crystallinity with negligible PbI2 formation, in sharp contrast to the degraded control counterpart (Supplementary Figs. 33 and 34). These collective results unequivocally demonstrate that the self-assembly-directed, ordered crystallization induced by Cz-SCN is a highly effective strategy for producing n-i-p solar cells with the exceptional efficiency and long-term stability required for practical applications.

Discussion

In summary, we develop a highly effective strategy for directing the ordered crystallization of α-FAPbI3 in the conventional n-i-p architecture by incorporating a SAM, Cz-SCN, into the anti-solvent. Through a systematic comparison of anchoring groups, we demonstrate that the thiocyanate (-SCN) functionality is unique in its ability to drive the in-situ formation of a self-assembled guiding layer, a capability not observed for its analogues. Our synergistic investigation reveal that this capability stems from the superior adsorption energy of the -SCN group, which establishes a robust template on the perovskite surface. This template layer accelerates the efficient δ-to-α phase transition and guides the top-down crystallization process, thereby obtaining a highly ordered α-phase film with reduced defect densities and optimized energy level alignment for efficient hole extraction. This self-assembly-directed crystallization strategy translates directly into a profound enhancement in device performance and stability, yielding a champion PCE of 26.1% (certified at 25.67%) with outstanding operational and environmental stability. This work not only provides a powerful solution to the long-standing challenge of FAPbI3 crystallization control in n-i-p devices but also underscores the critical role of molecular design in guiding self-assembly for the advancement of next-generation photovoltaic technologies.

Methods

Materials

The reagents for Cz-SCN, Cz-COOH, Cz-CN, or Cz-NH2 synthesis were purchased from Aladdin and Energy Chemical (China). Formamidinium iodide (FAI) was sourced from Greatcell (Australia), while other reagents such as methylammonium chloride (MACl), cesium iodide (CsI), methylammonium bromide (MABr), lead bromide (PbBr2), Spiro-OMeTAD, lithium bis(trifluoromethanesulfonyl)imide (Li-TFSI), and FK209-Co(III)-TFSI were obtained from Xi’an Yuri Solar Co. (China). Lead iodide (PbI2) was supplied by TCI Shanghai (China). Solvents, including N,N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), chlorobenzene (CB), ethyl acetate (EA), isopropanol (IPA), acetonitrile (ACN), and 4-tert-butylpyridine (TB), were purchased from Thermo Fisher Scientific (USA). Stannous chloride (SnCl2·2H2O), thioglycolic acid (TGA), and urea were acquired from Sigma-Aldrich (USA). All reagents had a purity of at least 99%.

Device fabrication

The glass substrates with pre-patterned fluorine-doped tin oxide (FTO) were thoroughly cleaned. Ultrasonication was employed, with the substrates being successively immersed in detergent, deionized (DI) water, and ethanol, each immersion lasting for 20 min. After that, prior to utilization, the purified FTO-coated glass underwent a 15 min UV-ozone treatment. The electron-transporting layer consisted of a dense SnO2 film. This film was deposited on the FTO layer via a chemical bath deposition (CBD) method. The CBD solution was formulated as follows: per 100 mL of deionized (DI) water, 1.25 g of urea, 1.25 mL of HCl, 25 μL of TGA, and 275 mg of SnCl2 2H2O were mixed. The FTO-coated glass was then submerged in a diluted version of this mixed solution at 90 °C for 3 h. Subsequently, it was cleaned with DI water and IPA through sonication for 5 min, and thereafter annealed at 170 °C for 1 h. A 1.5 M perovskite precursor, FA0.98MA0.02Pb(I0.98Br0.02)3, was prepared. To achieve this, 252.79 mg of FAI, 711.56 mg of PbI2, 33.76 mg of MACl, 3.36 mg of MABr, and 11.01 mg of PbBr2 salts were dissolved in 1 mL of a DMF/DMSO (8:1, v/v) mixed solvent. A volume of 50 μL of the perovskite precursor solution was then dropped onto the CBD - SnO2 film. Consecutive spin-coating was carried out, first at 1000 rpm for 10 s and then at 5000 rpm for 30 s. During the last 10 s of spin-coating, 120 μL of ethyl acetate (EA) was added as an anti-solvent to the center of the perovskite film. The Cz-SCN, Cz-COOH, Cz-CN, or Cz-NH2 solution (3 mg/mL in EA) was then applied. The formed FAPbI3 film was annealed at 100 °C for 60 min. After this, a Spiro-OMeTAD layer was prepared. It had a concentration of 0.059 mmol mL⁻¹ in chlorobenzene (CB) and was doped with Li-TFSI (0.033 mmol), 4-tert-butyl pyridine (0.205 mmol), and FK209-Co (III)-TFSI (0.0057 mmol). This layer was spin-coated onto the perovskite film at 3000 rpm for 30 s. Finally, an 80 nm gold electrode was thermally evaporated onto the film, followed by a 100 nm MgF₂ coating on the glass. A perovskite module composed of six series-connected sub-cells was constructed on a 5 cm × 5 cm glass/FTO substrate. The interconnection were carried out by laser scribing at a wavelength of 532 nm. The active area of the module is 14.21 cm2. The geometric fill factor (GFF) calculation result is GFF = (4.0 cm−5 × 680 μm)/4.0 cm = 91.4%. The line widths of P1, P2, and P3 were 30 μm, 200 μm, and 150 μm, respectively, with inter-line spacings of 100 μm (P1-P2) and 200 μm (P2-P3).

Material characterization

1H-NMR and 13C-NMR were conducted in d6-DMSO using a Bruker 400 MHz instrument. High-resolution mass spectrometry (HRMS) tests are performed by Thermo Scientific Q Exactive Orbitrap instruments to provide accurate molecular mass measurements. Fourier Transform Infrared Spectroscopy (FTIR) measurement was performed by the INVENIO S Fourier Transform Infrared (FTIR). Measuring 64 scans per sample, from 4000 to 400 cm−1 with a resolution of 2 cm−1. Thermogravimetric analysis (TGA) was performed on Shimadzu DTG-60H at a heating rate of 10 °C min−1 under a nitrogen atmosphere, respectively. The absorbance spectra were measured by a UV–vis spectrophotometer with an integrating sphere (PerkinElmer, Lambda 950). 1H-NMR, 13C-NMR spectra, HRMS tests, and TGA characterization of Cz-molecules were showed in Supplementary Figs. 35–53.

Film characterization

SEM measurement was conducted using the JEOL JSM-7800F. Grazing-incidence wide-angle X-ray scattering (GIWAXS) experiments were conducted at the BL14B1 beamline of the Shanghai Synchrotron Radiation Facility, with samples prepared following the same film preparation procedure in the SSRF chemistry lab. Time of flight secondary ion mass spectrometry (ToF-SIMS) measurement. The instrument used was the Focused Ion Beam-scanning electron microscope at Westlake University. The in-situ PL spectra were recorded using an ISPL-HI001 instrument equipped with a 445 nm laser source (Nanjing Ouyi Optoelectronics Technology). In situ UV–vis absorption spectra were recorded using a setup where Y-type optical fibers were connected to the glovebox through a flange interface. A Hamamatsu EQ-99-FC served as the white light excitation source positioned outside the glovebox, while an Ocean Optics QE-PRO spectrometer was employed to monitor the real-time absorption changes during both the spin-coating and annealing stages. The surface morphology of the perovskite film was analyzed using AFM with the Park XE7 instrument. XPS measurements using the Thermo-Fisher ESCALAB 250Xi system equipped with a monochromatized Al Kα X-ray source were utilized. X-ray diffraction (XRD) measurements were acquired using a Smartlab 3KW diffractometer (Rigaku Corporation, Japan). Space charge limited current (SCLC) measurements were performed using the Keithley 2450 SMU. Time-resolved photoluminescence (TRPL) measurements were obtained using the Delta Flex TCSPC system from Horiba, with data collected at 820 nm under 530 nm excitation. UPS measurements were conducted using the PHI 5000 Versa Probe III instrument, equipped with a He I source (21.22 eV).

Device characterization

The current density-voltage (J-V) characteristics of the devices were measured by a Keithley 2400 source meter under a Class AAA sunlight simulator (XES-40S3, SAN-EI). The light intensity was calibrated to 100 mW cm−2 by a National Institute of Metrology (NIM) certified standard silicon reference cell (QEB1) at AM 1.5 G. Incident photon to electron efficiency (IPCE) testing was accomplished using a QER-900AD system (Nanjing Ouyi Photonics Technology) with a NIM-calibrated Hamamatsu Model S1337-1010BQ silicon detector for response correction. EIS and Mott-Schottky analysis were accomplished with the aid of a CHI760E electrochemical workstation (Chenhua Instruments, China). SPO tests were evaluated by biasing the device at the maximum power point (MPP) determined by its J-V curve and recording the steady-state current; the steady-state photoelectric conversion efficiency was calculated from the product of the applied voltage and the stabilized current density. Device humidity stability testing was conducted at 23 ± 2 °C and 40 ± 10% relative humidity, during which a Class AAA sunlight simulator was used to periodically monitor performance changes. In addition, MPPT tests were conducted under 1 sun irradiation in a nitrogen atmosphere at 20°C ± 3°C using a self-assembled multichannel potentiostat to further evaluate the device stability under operating conditions. For these measurements, we employed poly[bis(4‑phenyl)(2,4,6‑trimethylphenyl)amine] (PTAA) doped with 4‑isopropyl‑4′‑methyldiphenyliodonium tetrakis(pentafluorophenyl)borate (TPFB) instead of spiro‑OMeTAD, with a PTAA concentration of 30 mg mL−1 and a PTAA:TPFB weight ratio of 10:1.

Reporting summary

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