Enhancing Efficiency and Stability of Perovskite Solar Cells Through Electron‐Rich Covalent Organic Frameworks Radicals

ABSTRACT

Effective regulation of perovskite crystallization is crucial for achieving high‐performance perovskite solar cells (PSCs). However, perovskite films typically exhibit low crystallinity and are plagued by abundant bulk and grain boundary defects. Simultaneously controlling crystallization kinetics, defect passivation, and energy level alignment remains a significant challenge. In this study, we designed and synthesized an electron‐rich covalent organic framework (COF_FAT) and further introduced N‐cationic radicals (COF_Rad) within its framework and pores via a one‐step post‐treatment. The introduction of these radicals significantly reduced the COFs bandgap, enhanced charge transfer, and minimized open‐circuit voltage (V_OC) loss. The ordered COFs structure, featuring multiple coordination sites (Ph–N and N•⁺), modulated the crystallization process and effectively passivated bulk and grain boundary defects, thereby improving the crystallinity of α‐perovskite. As a result, PSCs incorporating COF_Rad achieved a remarkable power conversion efficiency (PCE) of 26.33% (certified 25.98%). These devices retained 88% of their initial PCE after 1000 h of thermal aging at 85°C, demonstrating outstanding durability. Moreover, COFs‐based PSCs exhibited excellent stability under continuous illumination and humid conditions. This work delivers the highest efficiency reported for COFs‐based PSCs to date and offers a new strategy for developing high‐performance and stable optoelectronic devices.

Introducing free radical–based covalent organic frameworks (COFs) as dopants into perovskite films enables effective modulation of the crystallization process and suppression of defect formation via interfacial passivation and electronic coupling. The incorporation of COFs enhances crystallinity, facilitates charge transport, and optimizes energy‐level alignment, thereby improving power conversion efficiency while markedly strengthening both operational and environmental stability.

1. Introduction

Over the past decade, the rapid enhancement of power conversion efficiency (PCE) and low cost have made perovskite solar cells (PSCs) one of the leading alternatives in photovoltaic technology [1, 2, 3]. The rapid increase in the PCE of single‐junction PSCs is largely attributed to the optimization of perovskite crystallization kinetics, which suppresses crystallographic defects [4, 5, 6]. As the crystal quality governs the photophysical properties of perovskite films, it critically influences device performance [7, 8, 9]. However, secondary phases (e.g., yellow δ‐phase) often form during crystallization, coexisting with or replacing the target black α‐phase (cubic/tetragonal) [10, 11]. Precise control over crystal microstructure, especially face orientation and growth dynamics, is essential for achieving high‐quality films with superior optoelectronic properties [12, 13].

The rapid and uncontrolled growth of perovskite crystals leads to numerous crystallographic defects [14]. These defects serve as non‐radiative recombination centers, hindering charge transport and reducing carrier lifetime and device performance [15]. Various strategies have been utilized to control the perovskite crystallization process [16]. Among them, additive engineering is widely regarded as a key strategy for passivating perovskite defects and controlling the crystallization process [17]. Organic small molecules [18, 19, 20], polymers [21, 22, 23], ionic liquids [24, 25, 26], and organic functional frameworks contribute to the passivation of uncoordinated lead ions (Pb²⁺) and the regulation of perovskite crystallization behavior [27, 28, 29, 30, 31]. However, the limited number of functional groups in small molecules results in fewer coordination sites, leading to limited passivation capability [32]. Moreover, in addition to the regulation of the crystallization process, the management of perovskite defects and the tuning of energy levels also pose challenges for existing additives [33].

Two‐dimensional covalent organic frameworks (2D COFs), as crystalline polymers with large surface areas, excellent thermal stability, tunable chemical composition, and photophysical properties, have attracted considerable attention in the field of photovoltaics [34, 35]. To date, numerous studies have incorporated COFs into PSCs to enhance device performance [36, 37, 38, 39]. The energy levels of COFs can be tuned by the chemical structures of the molecular building units, exhibiting excellent charge transport capabilities and facilitating efficient interfacial charge transfer in devices [38, 39, 40]. More importantly, due to the flexible designability of COFs, they can serve as bulk‐phase modifiers for PSCs, and may be directly incorporated into perovskite precursor solutions to promote crystal growth and improve the film quality [29, 36, 41, 42]. Meanwhile, radical‐containing materials have recently attracted increasing attention in photovoltaic and optoelectronic applications because of their unique open‐shell electronic structures and enhanced electronic delocalization [43, 44]. In particular, radical‐containing small molecules have been employed as PSCs additives to regulate energy levels and suppress non‐radiative recombination [45, 46]. However, these small‐molecule radical systems generally lack structural periodicity, extended electronic delocalization, and multiple coordination sites, limiting their ability to simultaneously regulate crystallization dynamics, defect passivation, and interfacial charge transfer. In contrast, radical‐functionalized COFs combine the advantages of periodic conjugated frameworks and stable delocalized radical structures, making them conceptually attractive as multifunctional interfacial regulators for PSCs. Nevertheless, radical‐containing COFs have not yet been explored in photovoltaic applications. More importantly, the relationship among radical electronic structure, crystallization kinetics, interfacial carrier dynamics, and device performance remains largely unclear. Therefore, developing radical‐functionalized COFs for synergistically regulating perovskite crystallization and interfacial electronic processes represents an important yet unexplored direction toward highly efficient and stable PSCs.

In this work, we designed and synthesized electron‐rich COFs (COF_FAT) and further constructed delocalized N‐cation radicals within the framework and pores through a one‐step post‐oxidation strategy, yielding a radical‐functionalized COFs (COF_Rad). By integrating the advantages of radical electronic delocalization and periodic COFs structures, COF_Rad simultaneously regulates perovskite crystallization, defect passivation, energy‐level alignment, and interfacial charge extraction [47, 48, 49]. Through in situ photoluminescence and grazing incidence wide‐angle x‐ray scattering (GIWAXS) characterizations, we directly reveal the role of radical COFs in accelerating α‐phase formation, suppressing impurity phases, and improving perovskite crystallinity. Furthermore, combined various characterizations, and theoretical calculations systematically elucidate the synergistic effects of radical electronic structures on interfacial energy‐level modulation, defect suppression, and carrier dynamics. Benefiting from these synergistic effects, the corresponding p–i–n PSCs achieve a champion PCE of 26.33% (certified as 25.98%) with negligible hysteresis, together with excellent operational stability. The COF_Rad‐based devices retain 88% of their initial efficiency after 1000 h of accelerated aging at 85° C and also exhibit remarkable stability under continuous illumination and humid conditions.

2. Results and Discussion

Our research began with the design of candidate electron‐rich unit molecules. These molecules can assemble into topologically ordered columnar π arrays that construct nanochannels for charge transport and serve as scaffolds through steric hindrance to guide the templated growth of perovskite [50]. To enhance hole transport and secure robust binding to the perovskite surface via Lewis acid‐base interactions, we integrated triaromatic amine, tribiphenylamine (TBPA) groups into the framework of the COFs (Scheme S1 and Figure S1). The TBPA moieties can strongly coordinate with free Pb²⁺ ions, thereby modulating the energy levels and defects of the perovskite while improving the crystallinity and stability of the film [51]. Inspired by this approach, we further explored 2D COFs, achieving the enrichment of triaromatic amine within both the framework and the pores. Figure 1a presents the design of COFs from scratch for device optimization. We selected N4,N4‐bis(4'‐amino‐[1,1'‐biphenyl]‐4‐yl)‐[1,1'‐biphenyl]‐4,4'‐diamine (TBA) and 4,4''‐bis(bis(9,9‐dimethyl‐9H‐fluoren‐2‐yl)amino)‐[1,1':4',1''‐terphenyl]‐2',5'‐dicarbaldehyde (FAT) as monomers, and successfully synthesized a 2D conjugated COF_FAT through an amine‐aldehyde condensation reaction. In this COF_FAT, triaromatic amine periodically distributed in the backbone and pores, enhancing the electron‐donating properties of the framework and thereby improving its hole transport ability. More importantly, we found that through a simple single electron transfer method, the nitrogen atoms in the triaromatic amine of COF_FAT can be oxidized to cationic free radicals. During this process, the free radical electrons delocalize throughout the framework and pores, forming a new free radical structure, COF_Rad (Figure 1a). Introducing free radicals into COFs, not only inserts additional carriers to the oxidizable COFs network, but also improves its conductivity and alters its electronic structure [52], providing a potential modifier for optimizing perovskite solar cell devices (Figure 1b).

FIGURE 1.

The design, synthesis, and characterization of COF_FAT and COF_Rad. (a) Synthetic routes of COF_FAT and COF_Rad. (b) Schematic diagram of the interaction and charge transfer between COFs and perovskites (not to scale). (c) EPR spectra of COF_FAT and COF_Rad powder. (d) DFT modeling of the interaction between the perovskite (100) plane with COF_FAT. (e) Pb 4f XPS spectra of control and COFs‐treated perovskite films. (f) I 3d XPS spectra of control and COFs‐treated perovskite films.

To understand the intrinsic electronic structure of COF_FAT, we modeled a representative COF_FAT fragment using density functional theory (DFT) (Figure S2). Calculation results show that the highest occupied molecular orbital (HOMO) of COF_FAT primarily spans the tribiphenylamine and fluorene moiety, while the lowest unoccupied molecular orbital (LUMO) is predominantly localized on the C═N bond and biphenyl moiety (Figure S3). Moreover, this electron‐cloud distribution is highly consistent with the blue areas that denote the electron‐rich zones in the electrostatic potential (ESP) map (Figure S4). UV‒vis absorption spectroscopy measurements reveal that after exposure to tetrafluoro‐1,4‐benzoquione (TFBQ), the absorption spectrum of COF_FAT exhibits a red shift, extending into the near‐infrared region (>750 nm) (Figure S5). This phenomenon is consistent with the expected formation of triarylamine radical cations. This may be due to the delocalization effect of the radical, which enhances the conjugated system and lowers the energy required for electron transitions, resulting in a red shift of the absorption band [52]. Correspondingly, the optical band gap of COF_FAT after oxidation decreases from 2.23 to 1.78 eV (for COF_Rad). Additionally, electron paramagnetic resonance (EPR) spectroscopy further validates this process, showing a strong radical signal, indicating that COF_Rad forms the radical cation (Figure 1c).

Crystal structure of COF_FAT was confirmed through powder x‐ray diffraction (PXRD) and simulation analysis. The PXRD pattern of COF_FAT exhibits distinct and well‐resolved diffraction peaks (Figure S6). Pawley refinement confirms an AA stacking mode, with optimized unit cell parameters of a = 45.678 Å, b = 45.678 Å, c = 5.826 Å, α = β = 90°, γ = 120°. The refinement residuals (Rwp = 4.16%, Rp = 2.80%) are both below 5%, indicating excellent agreement between the model and experimental data. The detailed crystal structure information and atomic coordinates are summarized in Table S1. In this structural model, N⁴,N⁴‐bis(4'‐amino‐[1,1'‐biphenyl]‐4‐yl)‐[1,1'‐biphenyl]‐4,4'‐diamine (ABT) and FAT molecules are periodically connected within the layers, forming a hexagonal grid; interlayer stacking occurs through π‒π interactions, forming one‐dimensional microporous nanochannels. The close interlayer π orbital overlap and continuous electronic channels significantly reduce the barrier for carrier transport, facilitating carrier movement. Notably, the COF_Rad retains a comparable degree of crystallinity to COF_FAT, with diffraction peaks at 2.15, 3.75, and 6.5° assigned to the (100), (110), and (300) planes, respectively (Figure S7). In Fourier transform infrared (FTIR) spectroscopy, the characteristic peaks representing amines and aldehydes disappear, and a new peak appears at ∼1610 cm⁻¹, attributed to the C═N stretching vibration mode (Figure S8). In addition, the N 1s x‐ray photoelectron spectroscopy (XPS) of COF_FAT and COF_Rad showed that the C═N related signals did not change significantly before and after oxidation, while signals attributed to N•⁺ radical cations appeared and the N 1s singnal of Ph‒N disappeared after oxidation, indicating that the oxidation reaction mainly occurred at the Ph–N site rather than C═N. The C 1s signal only exhibited a slight shift. Since the radicals are mainly formed around the nitrogen‐containing donor units, the variation in the C 1s peak position is relatively limited (Figure S9). Furthermore, the high‐resolution transmission electron microscopy (HRTEM) images distinctly display the well‐ordered lattice structure of the COFs, which is further validated by fast Fourier transform (FFT) analysis (Figure S10). These findings confirm the highly ordered crystalline structure of the COFs, consistent with the PXRD characterization results. At 77 K, the permanent porosity of COF_FAT and COF_Rad was determined via nitrogen adsorption isotherms. Meanwhile, the Brunauer‒Emmett‒Teller (BET) surface areas of COF_FAT and COF_Rad were 549.73 and 421.56 m² g^‒1, respectively (Figure S11). Although the BET surface area decreased after radical formation, the pore size distribution profiles of COF_FAT and COF_Rad remained generally similar. These characterization results indicate that the COFs maintained good crystallinity and specific surface area both before and after oxidation treatment. Moreover, after exposure to various solvents and heat treatments, the crystallinity of these COFs remained unchanged, demonstrating their excellent structural stability (Figure S12).

COF_FAT and COF_Rad were directly introduced into the perovskite precursor as functional additives to regulate the crystallization and energy levels of the perovskite (Figure 1b). The 2D COFs are rich electron‐donating units, particularly the radical COF_Rad, in which the unpaired electrons are delocalized throughout the framework and pores, enabling binding and bridging at interfaces and grain boundaries. DFT calculations were employed to validate the interaction between COF_FAT and the perovskite. As depicted in Figure 1d, the binding energy between COF_FAT and perovskite is –3.53 eV, indicating an interaction between COF_FAT and the perovskite due to the electron‐rich nature of the triaromatic amine and C═N units. To further assess the interaction between COF and perovskite, XPS was used to characterize the perovskite films before and after COF modification. Compared to the control, the Pb 4f core‐level peaks in the COF‐treated perovskite films shifted toward lower binding energies (Figure 1e), and similar shifts were observed for the I 3d peaks (Figure 1f). In addition, the N 1s XPS spectrum of the mixture of COFs and PbI₂ powders showed that the signal attributable to Ph‒N/C‒N•⁺ in COFs also shifted (Figure S13). This shift can be attributed to the interaction between the COFs and the halogenated lead species adjacent to the perovskite grain boundaries, thereby altering the electrostatic interactions between Pb²⁺ ions and the halide anions [53]. Small‐molecule model compounds further corroborate the interactions between the functional groups in the COFs (C═N, Ph–N, and N•⁺) and Pb²⁺/I– species (Figure S14‐15). EPR spectroscopy further confirms the role of radicals. In the perovskite precursor solution, the unpaired electrons of the N•⁺ centers in COF_Rad may partially participate in Pb∙∙∙N coordination or I–∙∙∙N⁺ electrostatic coupling, forming charge‐transfer or redistribution complexes that enhance spin pairing, accelerate spin relaxation, and thus lead to the attenuation of the EPR signal (Figure S16). Furthermore, DFT analysis indicates that the interaction between COF_FAT and perovskite increases the I vacancy defect formation energy (Figure S17), indicating that COF_FAT can passivate I vacancy defect. XRD patterns indicated the peak intensity of the perovskite (100) plane increases with the introduction of COF_FAT and COF_Rad, signifying an improvement in crystallinity (Figure S18). Notably, due to the trace‐level dispersion of COFs as dopants within the perovskite films, COFs signals are not detectable in the XRD spectra.

COF_FAT and COF_Rad are introduced as functional additives into the perovskite precursor to regulate the crystallization process (Figure 2a). To investigate the elemental depth profiles before and after treatment with COF_FAT and COF_Rad, time‐of‐flight secondary‐ion mass spectrometry (TOF‐SIMS) was performed (Figure S19). The experimental results show that PbI₃ ^– ions are evenly distributed in the perovskite film. After adding COF, a significantly enhanced C₄ ^– signal was detected, and its intensity was one order of magnitude higher than that of the control group, indicating that COFs were successfully introduced into the perovskite film [54]. Dynamic light scattering (DLS) analysis of precursor solutions with COFs reveals that all samples contain colloidal particles of approximately 50 nm (Figure S20). Notably, large aggregates (∼1 µm) appear in solutions containing COF_FAT and COF_Rad, which may be attributed to an intermediate phase formed by strong interactions between COFs’ functional groups and Pb²⁺/I^‒ ions. Such pre‐nucleation clusters can serve as nucleation sites, promoting rapid nucleation and inducing highly oriented grain growth.

FIGURE 2.

The dynamics of perovskite nucleation and crystallization. (a) Schematic illustration of the formation processes for the perovskite film before and after COFs modification (not to scale). (b‒d) In situ PL spectra of the (b) control, (c) COF_FAT‐, and (d) COF_Rad‐treated films during the spin‐coating process. (e‒g) In situ PL spectra of the (e) control, (f) COF_FAT‐, and (g) COF_Rad‐treated films during the annealing process.

In situ photoluminescence (PL) spectroscopy was used to monitor the effect of COFs addition on the nucleation and crystallization of COFs on perovskite during spin coating and thermal annealing. During the spin‐coating process of the control perovskite films, after adding the anti‐solvent, after 15 s the PL peak intensity of the control film increases at ∼800 nm (Figure 2b), indicating the nucleation of α‐perovskite. The peak intensity decreases after 70 s and rapidly decays within the following 10 s, indicating the formation of a photoluminescence quencher. In comparison, the PL peak intensity of the COFs‐treated film significantly increases, with no obvious decay, indicating a faster nucleation rate and less non‐radiative recombination (Figure 2c,d) [55]. This phenomenon may be attributed to the interaction between the functional groups (such as Ph‒N, N•⁺) in the ordered structure of COFs and Pb²⁺/I^‒ ions to form clusters, thereby accelerating the nucleation process, which is also consistent with the DLS measurement results.

During thermal annealing of the the control perovskite films (Figure 2e‒g), the formation of α‐perovskite leads to a rapid increase in PL intensity, which then sharply decreases due to solvent evaporation, crystal reconstruction, and the negative temperature coefficient effect [56]. Subsequently, the PL intensity recovers, indicating the recrystallization of α‐perovskite. The COF_FAT and COF_Rad‐treated films begin recrystallization after 34 and 42 s, respectively, with a significant delay compared to the control film (12 s), indicating slower crystal growth. This time delay may be due to the strong interaction between COFs and Pb²⁺/I^‒ ions, particularly the free radical electrons in COF_Rad, which inhibit the crystal formation of perovskite. Small molecule model compounds containing/without free radicals further demonstrate the role of free radicals (Figures S21 and S22). Rapid nucleation and slow crystal growth are key to obtaining high‐quality perovskite films [57]. The results indicate that COFs modification can effectively regulate crystal growth and suppress non‐radiative recombination.

The morphological changes of perovskite films before and after COFs modification were observed by using scanning electron microscopy (SEM) and atomic force microscopy (AFM). As Figure 3a‒c shown, compared to the control film, the COFs‐treated perovskite films exhibit no obvious defects, and the average grain size of the COF_FAT‐ and COF_Rad‐treated films increases from 0.39 to 0.47 and 0.49 µm, respectively. Cross‐sectional SEM images show that the COFs‐treated films display larger grain sizes and fewer grain boundaries (Figure S23). The formation of larger through‐grains minimizes grain boundaries, reducing transport barriers and enhancing vertical charge extraction. In addition, AFM images (Figure S24) show that the root mean square roughness (RMS) of the COF_FAT‐ and COF_Rad‐treated perovskite films is 18.2 and 16.7 nm, respectively, which are lower than the 22.1 nm of the control film. These results indicate that incorporating COFs facilitates the formation of smooth, uniform, and large‐grained perovskite films.

FIGURE 3.

Characterization of the control, COF_FAT‐ and COF_Rad‐treated perovskite films. (a‒c) Top‐view SEM images with inset showing the corresponding grain size distribution. (d‒f) The structural evolution monitored by azimuthally integrated in situ GIWAXS during heating from RT to 100 °C using the as‐deposited thin films. (g) HRTEM image of COF_Rad‐treated perovskite films, the inset is the SAED patterns. (h) The cross‐sectional cryo‐TEM image of the COF_Rad‐treated perovskite films. (i) The dextral lattice fringe images collected from the corresponding regions in (h). (j) The corresponding intensity profile in (i).

To further investigate the role of COFs in the crystallization process of Cs_0.05FA_0.95PbI₃ perovskite films from precursor materials, as shown in Figure 3d‒f, we monitored the structural evolution of the films deposited at room temperature when heated to 100 °C using in situ (GIWAXS). The δ‐Cs_0.05FA_0.95PbI₃ of the control perovskite film completely transformed into α‐Cs_0.05FA_0.95PbI₃ after heating for 30 s (Figure 3d). Notably, compared to the control sample, the COF_FAT‐treated perovskite film showed almost complete disappearance of δ‐Cs_0.05FA_0.95PbI₃ after only 19 s of heating. The COF_Rad‐treated perovskite film even only exhibited α‐Cs_0.05FA_0.95PbI₃ at room temperature, and no δ‐Cs_0.05FA_0.95PbI₃ was found (Figure 3f). These phenomena indicate that the introduction of COFs promotes the transformation of δ‐Cs_0.05FA_0.95PbI₃ to α‐Cs_0.05FA_0.95PbI₃ [58]. Additionally, the GIWAXS results show that after introducing COFs, the diffraction ring signal corresponding to the (100) crystal plane (q value of 1 Å⁻¹) of the film is brighter, indicating a significant enhancement in crystallinity (Figure S25).

HRTEM confirms the existence of COFs within the perovskite, revealing that the two‐dimensional layered COFs are mainly distributed at the grain boundaries (Figure S26), where lattice spacings of 11.25 and 3.18 Å correspond to the (220) plane of COF_Rad and the α‐phase perovskite, respectively. Furthermore, HRTEM images further verified the uniformity of COF_Rad‐based perovskite films. The clear lattice (lattice spacing of 3.18 Å) and single‐crystal selected area electron diffraction (SAED) patterns showed that the COF_Rad‐treated perovskite films were highly uniform (Figure 3g). To further evaluate the crystal uniformity of the top and bottom regions of the film, we used focused ion beam (FIB) technology. All TEM images of the COF_Rad‐treated films exhibited an ordered matrix structure with a 3.18 Å interplanar spacing (Figure 3h‒j). The measured d‐spacings for the COF_FAT‐based perovskite film were d _bulk = 3.18 Å and d _bottom = 3.15 Å, while those for the control perovskite film were d _bulk = 3.12  Å and d _bottom = 3.06 Å (Figure S27). The decreasing trend in d‐spacing at the film bottom indicates significant lattice mismatch, which may be associated with increased internal lattice stress in the perovskite layer [59]. Furthermore, HRTEM images show that the COF_Rad‐based film exhibits markedly improved crystallinity compared to the other two films.

Further analysis of carrier dynamics in perovskite films before and after COF modification was conducted by using femtosecond transient absorption (TA) spectroscopy. The 2D TA spectra show that the COF_Rad‐treated perovskite film presents less faded ground‐state bleaching (GSB) signal, as well as a stronger GSB intensity than that of the COF_FAT‐treated and control perovskite films (Figure 4a‒c). The lifetimes calculated by using a single‐exponential kinetic model were 1501.70 (control), 1723.14 (COF_FAT), and 2760.19 ps (COF_Rad), further confirming the reduction in defect density in COFs‐treated perovskite films (Figure S28).

FIGURE 4.

Effect of COF doping on the photoelectric properties of perovskite films. (a–c) 2D TA spectra of the (a) control, (b) COF_FAT, and (c) COF_Rad‐added perovskite films. (d) Steady‐state PL spectra of the control, COF_FAT‐ and COF_Rad‐treated perovskite films deposited on glass. (e) UPS spectra of the control and COFs‐treated perovskite films. (f) The energy levels of the VBM and CBM obtained from the UPS spectra for the control and COFs‐treated perovskite films.

The non‐radiative carrier recombination behavior of perovskite films before and after the introduction of COFs was evaluated through steady‐state PL and time‐resolved photoluminescence (TRPL) spectroscopy. The PL intensity of COFs‐treated perovskite films was enhanced (Figures 4d and S29). The biexponential‐fitted TRPL results indicate that the addition of COFs effectively suppresses fast non‐radiative decay (Figure S30). As the quality of the perovskite films improved, the photoluminescence lifetime increased from 452.68 (Control) to 722.72 (COF_FAT), and eventually to 2520.51 ns (COF_Rad, Table S2). The above results consistently indicate that COFs modification can suppress the non‐radiative recombination of carriers, thereby improving the carrier lifetime.

Bader charge analysis of the COF_FAT‐perovskite (100) plane was performed by using DFT calculations. As shown in Figure S31, an increase in electron density within the perovskite structure was observed, with 0.02e transferred from COF_FAT to the perovskite. This phenomenon can be attributed to the presence of electron‐rich triaromatic amine in the COFs structure, enhancing the overall electron cloud density of the framework. The resulting increase in electron density leads to a more favorable energy level alignment between the COFs‐treated perovskite layer and the electron transport layer. Furthermore, UV‒visible absorption spectroscopy characterization demonstrates that the modification by COFs has negligible impact on both the transparency of the substrate and the characteristic absorption region of the perovskite films (Figure S32).

Ultraviolet photoelectron spectroscopy (UPS) was employed to further analyze the energy band structure of the perovskite films. As shown in Figure 4e, the secondary electron cutoff (SECO) and conduction band maximum (CBM) spectra were measured. The results indicate that the fermi level (E _F) of the COF_FAT‐ and COF_Rad‐treated perovskite films increase from ‒4.10 to ‒4.06 and ‒4.02 eV, respectively, while the CBM shifts from ‒3.86 to ‒3.94 and ‒3.96 eV, compared to the control (Table S3). Kelvin probe force microscopy (KPFM) images show that, the contact potential difference (CPD) of the COF_FAT (‒360 mV) and COF_Rad‐treated perovskite films (‒250 mV) is significantly higher than that of the control group (‒410 mV), indicating a reduction in work function and an upward shift of the Fermi level, consistent with the UPS results (Figure S33). These energy level variations are illustrated in Figure 4f. Compared with COF_FAT treatment, the COF_Rad‐treated perovskite film exhibits a pronounced n‐type characterstic and band bending. The larger n‐type Fermi‐level shift induced by COF_Rad modification is believed to facilitate electron extraction and hole blocking, thereby effectively suppressing interfacial charge recombination [52].

To study the effect of COFs doping on film defects and carrier mobility, space charge limited current (SCLC) tests were performed (Figure S34). In electron‐only devices, the trap filling limit voltage (V _TFL) of devices treated with COF_FAT and COF_Rad was 0.092 and 0.089 V, respectively, lower than the 0.097 V of the control device, and the defect state density was 1.329 × 10¹⁵ (COF_FAT) and 1.285 × 10¹⁵ (COF_Rad) cm⁻³, respectively (the defect state density of the control film was 1.401 × 10¹⁵ cm⁻³). Hole‐only devices also showed similar trends, with V _TFL of 0.122 (COF_FAT), 0.103 (COF_Rad), and 0.140 V (Control), and defect state density of 1.762 × 10¹⁵ (COF_FAT), 1.488 × 10¹⁵ (COF_Rad), and 2.022 × 10¹⁵ cm⁻³ (Control). The results reveal that COFs incorporation reduces film defects and improves perovskite crystallinity, thereby facilitating long‐range carrier transport and enhancing vertical charge extraction. Moreover, the COF_Rad‐ treated film exhibits a maximum photoluminescence quantum yield (PLQY) of 23.35% and the highest quasi‐Fermi level splitting (QFLS) of 1.217 eV (Figure S35), which corresponds to the minimal open circuit voltage (V _OC) loss.

To evaluate the impact of COFs incorporation on device performance, we fabricated p–i–n PSCs (Figure 5a) with the following layer stack: ITO//Me‐4PACz/Cs_0.05FA_0.95PbI₃(COFs)/PCBM/BCP/Ag (Figure 5b). The optimal concentrations of COF_FAT and COF_Rad verified through experiments were both 0.005 mg mL⁻¹ (Figure S36 and Table S4). Under simulated AM 1.5G one‐sun illumination, we measured current density‒voltage (J–V) characteristics of devices with an active area of 0.09 cm² (Figure 5c). The control device exhibited, a V _OC of 1.185 V, a current density (J _SC) of 25.63 mA cm^‒2, and a fill factor (FF) of 79.22 % and a PCE of 24.06 %. Incorporation of COF_FAT increased V _OC to 1.193 V, J _SC to 25.88 mA cm^‒2, FF to 82.85 %, and PCE to 25.58 %. Notably, the COF_Rad‐based PSC achieved a champion PCE of 26.33 %, with V _OC of 1.206 V, J _SC of 25.94 mA cm⁻², and FF of 84.17 %, exhibiting negligible hysteresis (Table S5). The optimum‐performing device was independently certified at 25.98 % PCE by an accredited solar‐cell testing laboratory (Figure S37). The enhancements in V _OC and FF for the COFs‐treated devices are ascribed to improved perovskite crystallinity and suppressed nonradiative recombination losses. Importantly, for the PCE employing 2D organic functional materials including both COFs and metal–organic frameworks (MOFs), we have realized one of the highest reported performances to date for p–i–n PSCs, demonstrating the efficacy of our strategy (Table S6). External quantum efficiency (EQE) spectra of the COF_Rad devices yielded an integrated J _SC of 25.52 mA cm^‒2, in excellent agreement with J‒V measurements (25.24 and 25.01 mA cm^‒2 for COF_FAT and control, respectively). From the EQE onset, the perovskite bandgap was estimated at ∼1.53 eV, consistent with that determined from UV–vis absorption measurements (Figure S38). To assess reproducibility, we fabricated 15 independent devices for each condition; the resulting statistical distributions (Figures 5e and S38) confirm the robust performance of the COF‐treated PSCs. In addition, the universality of our radical COFs doping strategy was demonstrated across perovskite compositions and device architectures (Figure S40). Finally, steady‐state power output (SPO) measurements at the maximum power point over 300 s further validate the reliability of our devices (Figure 5f).

FIGURE 5.

Device structure and performance of the inverted PSC with COFs addition. (a) Device structure of the inverted PSC. (b) Cross‐section SEM of the inverted PSC. (c) J–V curves of p–i–n PSCs at laboratory scale. (d) EQE curves of the COF_Rad‐treated devices. (e) Efficiency statistics and (f) SPO curves of the control and COFs‐treated devices. (g) Dependence of V _oc on light intensity. (h) Nyquist plots and (i) Dark J–V curves. (j) Mott–Schottky plots. (k) Environmental stability at 20% relative humidity of controlled and COFs‐treated devices. (l) Light stability of unencapsulated controlled and COFs‐treated devices. m Thermal stability of controlled and COFs‐treated devices.

To elucidate the mechanism by which COF modification enhances device performance, we performed comprehensive optoelectronic analyses on the control and COFs‐treated PSCs. Figure 5g plots V _OC as a function of light intensity, in which the slopes extracted for the COF_FAT‐ and COF_Rad‐treated devices are 1.56 and 1.37 k _B T/q, respectively, and both data are significantly smaller than the 1.67 k _B T/q observed for the control devices. This reduction in slope indicates a suppressed density of trap‐assisted recombination in the COFs‐treated devices. To probe charge‐transfer dynamics, we carried out transient photovoltage (TPV) and transient photocurrent (TPC) measurements. TPV decay lifetimes for the control, COF_FAT, and COF_Rad devices are 14.50, 12.73, and 12.00 ms, respectively (Figure S41), suggesting that bulk defect densities in the COFs‐treated perovskite films are significantly reduced. Conversely, TPC decay lifetimes decrease from 1.11 µs in the control to 0.79 (COF_FAT) and 0.35 µs (COF_Rad), implying more efficient charge extraction at the perovskite/transport‐layer interfaces upon COF modification (Figure S42). This behavior is attributed not only to enhanced electronic delocalization and interfacial charge‐transfer capability enabled by the radical‐functionalized framework, but also to suppressed trap‐assisted recombination arising from effective defect passivation. Electrochemical impedance spectroscopy (EIS) further corroborates these findings (Figure 5h, Table S7): COFs‐based devices exhibit lower series resistance (R _s) and higher recombination resistance (R _rec), indicating that charge recombination is effectively inhibited, thereby facilitating improved charge transport. Dark J–V measurements (Figure 5i) reveal reduced leakage currents in COFs‐treated PSCs, demonstrating suppression of Shockley–Read–Hall (SRH) recombination pathways. Capacitance–voltage (C–V) profiling (Figure 5j) was used to extract the built‐in potential (V _bi). The COF_Rad device shows the highest V _bi of 1.05 V, compared to 1.02 V for COF_FAT and 0.96 V for the control, suggesting stronger internal electric fields and more effective carrier separation and transport—factors that underpin the observed enhancements in V _OC and FF.

The stability of the devices before and after COFs modification under various conditions was evaluated. In the humidity stability test (20% RH, in air, Figure 5k), the COF_Rad‐based PSC retained 96% of its initial PCE after 1000 h significantly outperforming both COF_FAT (93%) and the control device (83%). For photostability, under one‐sun white‐light LED illumination (100 mW cm^‒2) in N₂ (Figure 5l), the control device retained only 75% of its initial PCE after 1000 h, whereas the COF_FAT and COF_Rad‐treated devices retained 90% and 93%, respectively. The thermal stability of the unencapsulated devices at 85 °C in N₂ (Figure 5m) showed that the COF_Rad‐treated device maintained 88% of its original PCE after 1000 h, compared to 80% for COF_FAT and only 58% for the control. The COF_Rad‐treated device retained 94% of its initial PCE after 1000 h of continuous one‐sun illumination under maximum power point (MPP) tracking (Figure S43). To further investigate this difference, we systematically characterized the perovskite films with and without COFs treatment under 35% relative humidity and 65 °C conditions. The COF‐treated perovskite films exhibited enhanced stability, consistent with the device stability tests (Figure S44–S46). Furthermore, we measured the water contact angles of the devices. The contact angles of the COF_FAT‐ and COF_Rad‐treated films were 63.8° and 70.3°, respectively, both higher than the 58.7° observed for the control film, confirming that COFs treatment enhanced the hydrophobicity of the film (Figure S47). The improved device stability arises from the synergistic role of COFs at perovskite grain boundaries. Their ordered frameworks provide well‐defined nucleation sites that enhance crystal quality and film density, while structural embedding and Pb^{2 +}coordination effectively passivate defects, suppress non‐radiative recombination and ion migration, and stabilize the lattice. The aromatic‐rich backbone further strengthens film hydrophobicity, enhancing environmental durability. Together, these structural and chemical effects fundamentally reinforce the long‐term photo‐, thermal‐, and moisture stability of perovskite films and devices.

3. Conclusion

In summary, we have developed a novel electron‐rich COF_FAT and introduced N‐cationic radicals COF_Rad via a simple post‐treatment strategy, achieving the first successful integration of radical COFs in the PSCs. The presence of N•⁺ radical cations, not only narrows the bandgap of the COF and facilitates charge transfer but also improves the energy level alignment of the perovskite layer, effectively reducing the V_OC loss. The synergistic effects of the ordered COF structure and multiple coordination sites (Ph–N, N•⁺) enable regulation of the perovskite crystallization process and effective passivation of both bulk and grain boundary defects, leading to improved crystallinity of α‐perovskite. Benefiting from these effects, COF_Rad‐based PSCs achieved a champion PCE of 26.33% (certified 25.98%), along with remarkable operational stability, retaining 88% of their initial efficiency after 1000 h of thermal aging at 85 °C. The devices also showed excellent resistance to moisture and continuous illumination. This work lays the foundation for the design and development of COF‐based additives compatible with perovskite manufacturing processes, aimed at enhancing the structure, energy level alignment, and optoelectronic properties of PSCs.

Author Contributions

Shuai Yang: investigation, writing – original draft, writing – review and editing. Jiaxin Ma: investigation. Hao Luo: investigation. Jinbei Wei: investigation. Wei Huang: investigation. Tianhong Huang: investigation. Chang Cui: investigation. Jiadi Chen: investigation. Weifeng Zhang: investigation. Jizheng Wang: investigation, writing – review and editing. Gui Yu: Investigation, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: anie72921‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.