Introduction

Perovskite/silicon tandem solar cells (TSCs) are a highly promising technology that combines wide-bandgap (WBG) perovskites in the top subcells with commercially available crystalline silicon in the bottom subcells1,2. The recorded power conversion efficiency (PCE) of perovskite/silicon TSCs has reached 34.6%, offering promising potential for substantial reductions in the levelized cost of electricity (LCOE) for commercial solar panels. Perovskite/silicon TSCs can be fabricated using silicon wafers with flat, nanoscale-textured, or fully textured surfaces. Creating a flat wafer surface requires an additional polishing step, which notably increases manufacturing costs and results in substantial optical absorption losses due to higher reflectivity3,4,5,6,7,8,9. Although fully textured wafers with pyramid sizes ranging from 3−10 μm minimize optical loss, they pose challenges for the conformal deposition of high-quality perovskite films. This process, requiring thermal co-evaporation, not only increases equipment costs but also complicates the development of effective strategies to enhance film quality10,11,12,13,14,15. By contrast, nano-textured wafers with pyramid sizes below 1 μm exhibit exceptionally low reflection loss, comparable to fully textured surfaces, while also offering excellent compatibility with cost-effective solution deposition processes. Notably, the majority of efficient perovskite/silicon TSCs are currently fabricated using nano-textured silicon bottom subcells16,17,18,19,20,21.

Considering the submicron-scale pyramid structure on the silicon wafer, it is essential to prepare high-quality wide-bandgap (WBG) perovskite films with a thickness of 1 μm to ensure complete coverage up to the peaks of the pyramids. However, WBG perovskites with high bromine (Br) and cesium (Cs) ion content often face challenges in controlling crystallization dynamics due to the limited solubility of the precursors, resulting in degradation of both the crystallinity and morphology of the perovskite films22. The presence of a significant number of lateral and vertical grain boundaries in the 1-μm thick perovskite films accommodates a high density of trap states. This, in turn, leads to pronounced non-radiative recombination and halide migration, compromising both the photovoltaic performance and long-term stability of the devices, despite the excellent defect-tolerant properties conferred by the antibonding nature of the valence band23. Therefore, significant efforts have been dedicated to the fabrication of high-quality thick perovskite films on nanotextured silicon surfaces through extensive research on material compositions and technological processes.

Additive engineering has been established as a highly effective strategy for optimizing the comprehensive film-forming process, encompassing solution microstructure, crystallization kinetics, and film morphology. This approach facilitates the attainment of efficient and stable perovskite films characterized by significantly reduced defect density. Within the broad spectrum of extrinsic additives, conjugated molecules possessing a rigid planar structure and π-electron delocalization have garnered substantial interest due to their exceptional electrical conductivity and intermolecular π-stacking interactions24,25. The presence of a delocalized electron cloud within the conjugated structure enhances the interaction between their functional groups and the perovskite lattice26. Additionally, the π-stacking effect facilitates the formation of orderly alignment on perovskite polycrystallines27. These synergistic effects induce robust interfacial dipoles and mitigate interfacial disorder, while simultaneously enhancing the interactions between the perovskite and the contact layers through the conjugated groups. Moreover, the functional groups affixed to the conjugated core are particularly significant, as they engage directly with the perovskite lattice28. Amphoteric compounds, exhibiting both anionic and cationic functionalities, possess the capacity to effectively neutralize both negative and positive electrical defects within perovskite films. Consequently, it becomes feasible to design exceptional additive materials through the construction of amphoteric coplanar conjugated molecules (ACCMs). However, in ACCMs, the inductive effect between acidic and basic moieties is exacerbated by the conjugated effect, leading to substantial alterations in the density distribution of the electron cloud29. Such alterations in chemical properties may potentially attenuate the interaction between functional groups and perovskites or even render the additive inactive. Therefore, a more rational design of the structural framework for ACCMs is imperative to optimize their efficacy.

In this study, we commenced by utilizing benzoic acid as a foundational material and, through precise application of density functional theory (DFT) and Brønsted acid-base theory, rationally designed an amphoteric coplanar conjugated molecule (ACCM), specifically 2-methyl-3H-benzimidazole-5-carboxylic acid (hereafter referred to as MBC), as an effective additive. The imidazole moiety within MBC confers suitable basicity while concurrently enhancing the acidity of benzoic acid through an electron-withdrawing effect, thereby facilitating the deprotonation process. As a result, MBC can exist simultaneously as an intramolecular salt and carboxylate anion, which is advantageous for its dual role in modulating crystallization dynamics and passivating charged defects. Building upon this, we incorporated MBC into the Cs0.1FA0.65MA0.25Pb(I0.8Br0.2)3 (Eg = 1.67 eV) perovskite precursor solution to fabricate high-performance WBG PSCs. A series of spectroscopic analyses were conducted to elucidate the interaction between MBC and the perovskite matrix, complemented by in-situ experiments to assess the influence of MBC on crystallization kinetics. Additionally, we investigated the impact of MBC on the crystal structure and optoelectronic properties of the perovskite films, confirming its positive role in enhancing carrier extraction and reducing non-radiative recombination. With the incorporation of MBC, we successfully achieved high-quality perovskite films even with micron-level thicknesses. Ultimately, this approach enabled the fabrication of high-efficiency TSCs with outstanding long-term stability.

Results and discussion

Design concept of MBC

As previously discussed, ACCMs hold significant potential as additive materials. The carboxyl group, as an anionic moiety, has been demonstrated to possess exceptional capability in interacting with precursors and passivating charged defects30,31. It is important to highlight that the carboxyl group exhibits weak acidic properties, with its protonation to-COOH or deprotonation to -COO− being dependent on its dissociation constant. Moreover, the passivation mechanisms of the carboxyl (-COOH) and carboxylate anion (-COO−) differ, a distinction that has often been overlooked in existing studies. The -COO− species tends to form ionic interactions with positively charged defects, while the -COOH species should engage in hydrogen bonding with negatively charged undercoordinated halides. The former may be significant due to the tendency of the perovskite surface to adopt lead iodide as the terminal, accompanied by a substantial number of halogen vacancies and undercoordinated lead ions. We use benzoic acid as a template to compare the interaction strengths between its anionic form and neutral form with PbI2. As shown in Fig. 1a, the negative charge of the benzoic acid anion directly interacts with the Pb2+, while the carboxyl hydrogen and oxygen of benzoic acid interact with the I- and Pb2+ of PbI2, respectively. The binding energy of the benzoic acid anion with PbI2 is −3.86 eV, nearly double that of the neutral benzoic acid molecule with PbI2 (−2.04 eV). Consequently, in addition to targeting distinct defect types for passivation, carboxylic groups and their corresponding carboxylate anions display varying effectiveness in regulating crystallization dynamics, with the carboxylate anion proving more efficient. Notably, employing strong alkaline carboxylate salts in place of carboxylic acids to introduce carboxylate anions may lead to the deprotonation of organic cations by these alkaline additives, potentially causing rapid degradation of the precursor solution. Consequently, enhancing the acidity of carboxylic acids presents the most effective strategy for generating carboxylate anions. Videlicet, in designing amphoteric compounds based on benzoic acid, it is crucial to further increase the acidity of the carboxyl group, specifically by reducing its acidity coefficient (pKa).

Fig. 1: Molecular design and theoretical mechanism of MBC.
figure 1

a The chemical structure of benzoic acid (core), p-aminobenzoic acid (lower left) and MBC (lower right); the interaction between PbI2 and benzoic acid (upper left), as well as the benzoic acid anion (upper right). The red value corresponds to the pKa of the functional group, while the blue value represents the pKa of its conjugated acid. b The ESP images and calculated dipole moments of MBC, MBC- and MBC+-. c The calculated binding energies between MBC and several interacting species. The effect of MBC on the (d) defect formation energy and (e) DOS of various defects.

Full size image

In designing the cationic component for benzoic acid to form ACCMs, an amine group is typically introduced due to its ability to accept a proton and form an ammonium cation. However, according to the Brønsted-Lowry acid–base theory, amine groups are generally electron-donating, which can significantly reduce the acidity of the carboxyl group and hinder proton release, as evidenced by an increase in pKa from 4.08 to 4.77. Meanwhile, the pKa of the -NH2 group is only 2.69 (corresponding to the pKa of its conjugate acid, -NH3+), indicating its very weak basicity. Therefore, it is essential to incorporate a basic group with distinct electron-withdrawing capability when designing ACCMs. Imidazole with a conjugated planar structure has also been demonstrated to be an effective additive due to its amphoteric characteristics, enabling simultaneous interactions with lead halides and organic cations31,32,33,34. In addition, the nitrogen atoms within the imidazole ring exhibit electron-withdrawing properties, thereby reducing the electron density of the adjacent benzene ring. Wang et al. demonstrated that the imidazole group present in theophylline is essential for inducing molecular passivation by effectively forming hydrogen bonds with iodide ions through its active hydrogen, thus increasing absorption energy35. Given this finding, we attempt to combine imidazole with benzoic acid to form an ACCM, i.e., MBC, and subsequently evaluate the pKa of the functional groups. Consequently, MBC exhibits amphoteric properties, with the carboxylic acid groups functioning as proton donors, while the lone pair of electrons on the imidazole nitrogen serves as a proton acceptor. As per Fig. 1a, the pKa of -COOH and the 3rd N on the imidazole ring are 2.99 and 6.50, respectively. The introduction of imidazole not only enables the construction of ACCM, but also lowers the pKa of -COOH, which promotes the deprotonation process. Considering the amphoteric properties, Supplementary Fig. 1 illustrates the different forms of MBC present in solutions at various pH levels. The MBC molecule does not exist in its molecular form; instead, it exists as an imidazolon cation (named as MBC+) at pH levels below 2.99. Between pH 2.99 and 6.50, it predominantly exists as an internal salt (named as MBC+−). At pH levels between 6.50 and 12.79, it mainly exists as a carboxylic anion (named as MBC-), transitioning into a negatively charged bivalent anion (imidazole and carboxyl simultaneously losing hydrogen ions, named as MBC2−) as the pH continues to increase. The pH of the MBC aqueous solution is ~6 (Supplementary Fig. 2a), which aligns with the simulated results. The perovskite solution exhibits a pH value of around 7 (Supplementary Fig. 2b), leading us to conclude that MBC molecules primarily exist in the form of MBC-, with a minor portion present as MBC+-. Therefore, we deduced that MBC, in its various forms, could serve as an effective additive for perovskites with complex compositions and diverse types of defects.

Theoretical impact mechanism of MBC

A range of DFT calculations were conducted to evaluate the property of MBC and predict its interactions with perovskites and contact layers. Figure 1b poses the molecular structure along with the electrostatic potential (ESP) map of MBC, MBC- and MBC+-. The carboxyl anion serves as the negative center, while the hydrogen atom of the imidazole ring acts as the positive center. In comparison to MBC, both the positive and negative potentials of MBC- and MBC+- exhibited significant increases, with the dipole moment rising from 5.4 D−16.9 D and 19.2 D, respectively. It is anticipated that these ionic species, characterized by substantial dipole moments, not only demonstrate a diverse range of strong interactions with perovskite precursors in solution but also effectively passivate various charged defects on the perovskite surface while simultaneously inducing interfacial dipole effects. Figure 1c and Supplementary Fig. 3 compare the binding energies of MBC- and DMSO to FA+ and PbI2. MBC- exhibits a stronger binding ability compared to DMSO, suggesting its potential to effectively modulate the crystallization dynamics. Moreover, the calculated π-stacking energies between MBC/MBC and MBC/SAM were found to be −0.09 eV and −0.61 eV, respectively, with a molecular plane distance of ~3.5-4 Å. Note that [4-(3,6-Dimethyl-9H-carbazol-9-yl) butyl] phosphonic acid (Me-4PACz) was utilized as SAM in this work. Both the energy values and molecular distances are consistent with the characteristic features of π-stacking36,37,38. These results indicate that MBC not only interacts with the perovskite precursor, altering crystallization kinetics, but also aligns on the crystal surface, thereby reducing interfacial energy disorder and promoting the formation of strong dipoles. Additionally, it will interact with the SAM molecules to facilitate charge extraction.

The solution-processed perovskite crystals, exhibiting apparent ionic characteristics, give rise to various types of charged defects on the surface of the films and at the grain boundaries. These defects, particularly the deep-level defects, lead to non-radiative recombination by annihilating electrons and holes, as described by the Shockley-Read-Hall (SRH) theory. The defects will also result in ion migration, thereby impacting the photovoltaic performance and stability of PSCs. Thus it is important to de-activate and reduce the density of deep charge traps through electronic passivation. The presence of Pb and I related point defects plays a pivotal role in determining the properties of perovskites, as the band edge of perovskites is dictated by the orbital characteristics of Pb and I. The capability of MBC in regulating the formation energy of defects and passivating point defects was evaluated through DFT calculations (Supplementary Fig. 4). The defect formation energies of Pb vacancy (VPb), I vacancy (VI), Pb-I antisite (PbI, substituting one Pb with one I atom) and I-Pb antisite (IPb, substituting one I with one Pb atom) were initially calculated. According to Fig. 1d, the presence of MBC leads to a significant increase in the defect formation energy for each defect, suggesting that the density of these defects in perovskites will be reduced when utilizing MBC as additives. Furthermore, the capability of MBC in passivating these defects was examined. As depicted in Fig. 1e, the VI represents a typical shallow-level defect. It should be noted that the role of shallow-level defects in perovskites remains a topic of ongoing debate. For instance, Guo et al. attributed the long carrier recombination lifetime in the (FA, MA, Cs)Pb(I1-xBrx)3 films to the presence of shallow states. These states were associated with a predominantly well-intermixed halide distribution, which transformed into deep traps due to light-induced phase segregation39. Zhou et al. demonstrated that the high-density shallow trap can temporarily hold one type of charges and increase the density of the other type of charges by keeping them from bimolecular recombination, thus reducing the voltage loss in PSCs40. On the other hand, recent studies by Yuan et al. have concluded that shallow defects predominantly govern the recombination process and remain a critical factor limiting device performance41. Moreover, Kim et al. have demonstrated that the shallow-level VI can generate polarons, triggering non-radiative processes and ultimately degrading the photovoltaic performance of PSCs42. The relationship between shallow defects and device performance, especially long-term stability, deserves further investigation in the future. While as per Fig. 1e, the presence of MBC- effectively eliminates the micro bulges in the density of state (DOS), indicative of passivation of VI. The VPb defect is identified as a deep-level defect with a high DOS near the center of the bandgap. Upon passivation with MBC, its intensity diminishes, and it shifts closer to the conduction band minimum. The PbI and IPb defects are recognized as deep-level defects positioned near the conduction band minimum and valence band maximum, respectively. Effective passivation of these defects can be achieved with MBC-, respectively, leading to the elimination or reduction of the DOSs. Therefore, owing to its amphiphilic nature, MBC simultaneously increases the defect formation energy and passivates both shallow- and deep-level defects, thereby holding significant potential to enhance the optoelectronic properties of perovskite films, although the precise role of shallow-level defects remains unclear.

Impact of MBC on the crystallization process

Given the anticipated substantial interaction between MBC and perovskite precursors, along with its potential to ameliorate defect density and passivate defects, we incorporated MBC into the perovskite precursors and systematically explored its impact on the solution microstructure, crystallization dynamics, and properties of resulting perovskite films. Both steady-state photoluminescence (PL) and UV-visible absorption spectroscopy were employed in order to investigate the interaction between MBC and perovskite in solution. Supplementary Fig. 5a demonstrates that neither MBC nor formamidinium iodide (FAI)/isopropanol (IPA) solution exhibits fluorescence within the 500-700 nm range. However, upon addition of MBC to the FAI/IPA solution, a new emission peak emerges at 550 nm, indicating the formation of a complex. As shown in Supplementary Fig. 5b, the absorption edges of FAI and MBC appear at 270 nm and 305 nm, respectively. Notably, their mixture exhibits a redshifted absorption edge at 310 nm, indicating potential charge transfer interactions between FAI and MBC. We subsequently examined the impact of MBC on the interaction between FAI and PbI2. The pure PbI2 in dimethyl formamide (DMF) solution exhibits absorption peaks at 285 nm, attributed to [PbI]+, and at 320 nm, corresponding to [PbI2]43,44. The introduction of FAI leads to the formation of the [PbI3]− species, which is evidenced by the absorption peak observed at 373 nm. However, the subsequent addition of MBC disrupts the presence of [PbI3]−, suggesting an interruption in the interaction between PbI2 and iodide ions (I−) originating from FAI (Supplementary Fig. 5c). This observation indicates that MBC exhibits a stronger binding affinity for PbI2 than for I−, highlighting its role in modulating the interactions within the precursor solution. In addition, both the Fourier transform infrared spectroscopy (FTIR) and nuclear magnetic resonance (1H NMR) spectroscopy also confirm the interplay between MBC and FAI and PbI2 (Supplementary Note 1).

In further, dynamic light scattering (DLS) measurements were performed to investigate the impact of MBC on the colloidal size in the perovskite precursor solution. As per Supplementary Fig. 9, both the control and MBC-incorporated (target) solutions display a unique bimodal distribution, spanning both micron and nanoscale particle sizes. The addition of MBC results in a reduction in the colloidal size across both scales. This reduction aligns with prior findings, where the addition of acidic components can decrease colloidal size45,46. But in reality, MBC incorporation increases the prevalence of larger colloidal particles while simultaneously reducing the number of smaller ones. Therefore, in terms of the average particle size, MBC leads to an increase in the colloid size. This implies that MBC facilitates the aggregation of smaller particles into larger structures, which may contribute to reduced nucleation density and enlarged grain size in perovskite films. To reveal the effect of MBC on the crystallization behavior of perovskite films, XRD patterns for as-coated wet perovskite films are depicted in Supplementary Fig. 10. The observed peaks at 6.29°, 6.89°, and 8.88° in the control sample are ascribed to the intermediate phases encompassing various solvates, while the peak at 13.8° corresponds to the perovskite phase47. These intermediate phases may lead to the formation of imperfections in the perovskite films, including defects, secondary phases, voids, and pinholes. The solvate peaks in the target sample, on the other hand, exhibit lower intensity, indicating that MBC inhibits the formation of the intermediate phase. It is worth noting that, the WBG perovskite wet films with a composition of Cs0.1FA0.65MA0.25Pb(I0.8Br0.2)3 does not involve δ phase due to their high Cs and Br content. To precisely investigate the impact of MBC on the crystallization dynamics, we conducted in-situ UV-visible absorption measurements on anti-solvent-treated wet films heated at 70 °C with a temporal resolution of 0.2 s. As shown in Fig. 2a, b, the crystallization rate of the target perovskite film is significantly slower than that of the control sample. This observation is further corroborated by Supplementary Fig. 11, which presents the time-dependent absorption intensity at 550 nm. The control and target samples reached their respective absorption plateaus at 33 s and 79 s, respectively. The reduction in solvate formation and the retardation of the crystallization process can both be attributed to the strong interaction between MBC and the perovskite precursor, as clearly demonstrated by the results presented in Fig. 1c.

Fig. 2: MBC regulates crystallization kinetics and film morphology.
figure 2

The in-situ UV-visible absorption spectra of perovskite films (a) without and (b) with MBC during annealing at 70 °C. c The XRD patterns of perovskite films with MBC of various concentrations. d The surface energy (γ) of various facets and the absorption energy (Ea) of MBC on various facets. e The calculated texture coefficient of various facets of perovskite films. f Radially integrated intensity plots along the ring at q = 19 nm-1 from GIWAXS maps. g, h The top-view and cross-sectional SEM images of perovskite films. i The conductive AFM images of perovskite films.

Full size image

The regulation of the crystallization dynamic process in the target sample is expected to enhance the quality of the final perovskite films. Figure 2c presents the XRD patterns with labeled Miller index of the control and target perovskite films. The preferred crystal facet of (100) tends to form spontaneously due to its lowest surface energy (γ) (Fig. 2d), which is in accordance with literature48. The incorporation of MBC increases the diffraction intensity of all crystal faces. Figure 2e shows the texture coefficient to quantitatively analyze the preferred growth of different crystal planes (Supplementary Note 2). By taking the scraped powder of the control sample as a reference, texture coefficients for (100), (110), and (111) planes were determined to be 1.15, 1.40, and 0.87 respectively in the control sample49,50,51. In contrast, the corresponding texture coefficients increased for (100) while decreasing for (110) and remaining unchanged for (111) in the target sample. The more favorable growth of the (100) facet in the target sample is attributed to the superior adsorption capability exhibited by MBC molecules on this specific facet (Fig. 2d). This finding is advantageous as the carrier mobility, exciton binding energy, and thermodynamic stability of the (100) and (111) crystal facets exhibit superior performance compared to that of the (110) facet52,53,54.

The top-view and cross-sectional scanning electron microscopy (SEM) images of the control and target perovskite films are depicted in Fig. 2g, h. In addition to demonstrating uniform coverage, a noticeable increase in the average grain size from 317−405 nm is observed. Moreover, the incorporation of MBC leads to a reduction in root mean square roughness (RMS), decreasing it from 27.6 nm to 23.9 nm (Supplementary Fig. 12). Both the enhanced grain size and reduced RMS imply a decrease in grain boundaries, which is advantageous for minimizing trap-state density. Interestingly, the cross-sectional SEM image of the target sample reveals an inclined texture, implying that the perovskite crystals exhibit a certain degree of preferential orientation. Supplementary Fig. 13 shows the grazing incidence wide-angle X-ray scattering (GIWAXS) results to further examine the effect of MBC on crystal growth orientation. The radially integrated intensity plots along the diffusion rings at q = 19 nm-1 (corresponding to the (200) plane) indicate that the target sample exhibits lower intensity at the azimuth angle of 50° while normalizing the intensity at 75°, suggesting a higher concentration of crystal growth along the out-plane direction in the target sample (Fig. 2f)55,56,57,58. This outcome is further confirmed by the conductive AFM results (Fig. 2i). The average surface tunneling current is increased from 0.152 nA to 0.313 nA due to the enhanced crystallinity and preferred growth orientation59, which will favor the charge carrier transport in perovskite films and the promotion of FF in PSCs60.

Effect of MBC on the property of perovskite

A series of spectroscopic analyses was conducted to assess the impact of MBC on the properties of the resulting perovskite films. The peaks in the X-ray photoelectron spectroscopy (XPS) of Pb 4 f orbit shift to higher binding energy because the electronegativity of MBC- is higher than that of I- (Supplementary Fig. 14a), as reflected by the more positive Mulliken charge of Pb cations in Pb (MBC-)2 than that in PbI2 (Supplementary Fig. 15). Moreover, the metallic Pb0 peaks at 136.6 and 141.6 eV in the control sample diminish after the introduction of MBC, indicating that the halide vacancy on the perovskite surface is passivated or eliminated. The peaks of I3d exhibits a slight simultaneous shift (Supplementary Fig. 14b), potentially due to modifications in the coordination environment of Pb2+ that affect the coupling of Pb-I bonds or the hydrogen bond interaction between MBC and I-. The time-of-flight secondary ion mass spectrometry (TOF-SIMS), as depicted in Supplementary Fig. 16, indicates that MBC molecules predominantly localize at the buried interface. This phenomenon can be attributed to the top-down crystallization direction of the perovskite, which results in the exclusion of MBC molecules as impurities. Supplementary Fig. 17 presents the SEM images of the buried surface of the perovskite films. The control sample reveals a significant presence of voids, including both small pinholes and larger cavities, with the latter measuring up to 200 nm in diameter. These voids are likely the result of capping layer formation during the rapid crystallization process, which traps solvents at the buried interface. Upon heating, the swift evaporation of solvents induces void formation. The presence of these voids hinders the hole transfer process and increases the series resistance in the devices. In contrast, the target sample exhibits a smoother buried surface with eliminated holes.

Furthermore, the grain size of the target is noticeably larger than that of the control, indicating a reduction in grain boundaries. Therefore, the target sample will impart enhanced hole transfer process and suppressed carrier recombination. In addition, the UPS measurements of the buried interface of the perovskite films, along with the extracted energy band diagram shown in Supplementary Fig. 18, demonstrate that MBC improves the band alignment at the buried interface by shifting the Fermi level upward and enhancing the p-type characteristics. This optimized band alignment reduces charge extraction barriers and facilitates more efficient charge transfer61. Subsequently, the Williamson-Hall (W-H) method was used to estimate the interface strain by measuring XRD as a function of the grazing incidence angles. As shown in Supplementary Fig. 19, the diffraction peaks shift toward lower angels as the incidence angles increase, signifying the presence of the tensile strain causing lattice expansion. However, the slope of the interplanar spacing of the (100) plane relative to the incidence angle in the control sample is steeper than in the target sample, indicating that interface strain is alleviated in the latter. The alleviated tensile strain may improve hole carrier transport and extraction by expanding the lattice, which broadens the bandgap and establishes an interfacial charge transfer barrier. Additionally, the release of strain contributes to lower defect density and enhanced long-term stability62.

Furthermore, a range of photoelectric measurements were utilized to investigate the effect of MBC on charge recombination, extraction, and ion migration processes. Figure 3a, b shows the transient absorption (TA) spectrum of perovskite films prepared on hole transport layers to gain insight into the hole extraction dynamics. The negative ΔA signal at 725 nm, identified as ground state bleaching (GSB), corresponds to charge depletion in the valence band of the perovskite. As shown in Fig. 3c, at a 6 ns timescale, the GSB signal of the target perovskite film decays more rapidly than that of the control, which is attributed to enhanced charge transfer from the perovskite to the SAM63. This improvement in hole extraction dynamics likely arises from π-π stacking interactions between MBC and SAMs. Supplementary Fig. 20a presents the steady-state photoluminescence (PL) of perovskite films prepared on SAM/NiOX/glass stacking. Regardless of whether excitation occurs from the buried surface or the top surface, the PL intensity of the target films exceeds that of the control, indicating suppressed non-radiative recombination and an enhanced degree of Fermi level splitting in the target. Figure 3d, e provides the steady state PL mapping of control and target perovskite films. The PL signal of the target exhibits a significantly enhanced intensity and demonstrates a more homogeneous distribution, implying an elevated level of surface component uniformity when compared to the control. Supplementary Fig. 20b depicts the time-resolved photoluminescence (TRPL) spectra of perovskite films, which consists of a long decay process associated with bulk recombination and a short decay process corresponding to surface recombination. The target exhibits a prolonged average lifetime compared to the control (758 ns versus 326 ns), as summarized in Supplementary Table 1. Furthermore, we utilized the double-heterostructure model to analyze the bulk lifetime and surface recombination velocity (SRV) by measuring the TRPLs of perovskite films with varying thicknesses (Supplementary Fig. 21)64. As shown in Fig. 3f, the target shows enhanced the bulk lifetime (1.26 μs versus 5.26 μs) and reduced SRV (67 cm/s versus 32 cm/s) compared with the control. These results confirm that the introduction of MBC enhances the quality of the film bulk and reduces the surface defect density65,66,67.

Fig. 3: Photoelectronic properties and ion migration suppression of perovskite films.
figure 3

a-b Pseudocolor TA plots. c Comparison of ground state bleaching decay kinetics at 725 nm of perovskite films with hole transport layers. d-e Steady-state PL mapping images of control and target perovskite films. f The fitting results of TRPL of varied thickness of perovskite films through a double-sided heterostructure model. g The temperature-dependent conductivity of control and target perovskite films. h-i The in-situ PL of control and target perovskite films.

Full size image

The space charge limiting current (SCLC) model was employed to evaluate the density of trap state and the carrier mobility of perovskite films (Supplementary Fig. 22, Supplementary Note 3)68,69. As summarized in Supplementary Table 2, the density of traps for electrons (holes) is calculated to be 1.9 × 1016 (1.7 × 1016) cm-3 and 1.1 × 1016 (0.9 × 1016) cm-3 for the control and target films, respectively. At the same time, the carrier mobility of the target is significantly higher than that of the control, with the mobility of electrons and holes being more balanced, which promotes efficient carrier transport and extraction. The increased carrier mobility is in excellent agreement with the C-AFM result. The ion migration is a critical phenomenon in perovskite materials due to the inherent characteristics of ionic crystals, which will generate deep-level defects, accelerate non-radiative recombination, exacerbate device hysteresis and cause irreversible degradation of films. In ABX3 perovskites, the migration activation energy of Pb2+ is the highest, followed by A-site organic cations, while X-site halide ions exhibit the lowest migration activation energy. Consequently, halide ions in perovskite exhibit the highest propensity for migration. Specifically, the migration of halides can induce phase segregation in WBG perovskite materials with a high concentration of bromide. The temperature-dependent conductivity of perovskite films was evaluated under dark conditions to measure the activation energy of ion migration (Ea). According to the Nernst-Einstein Eq. (1):

$$(\sigma )T=\frac{{\sigma }_{0}}{T}\exp \left(\frac{-{E}_{{{\rm{a}}}}}{{k}_{B}T}\right)$$
(1)

where σ0 is a constant, kB is the Boltzmann constant, T is the absolute temperature, and the Ea can be derived from the slope of the ln(σT)-1/kT relation. According to Fig. 3g, the Ea of control and target perovskite films is calculated as 171 meV and 283 meV, respectively. On this basis, the migration rate (r) is further calculated according to the Eq. (2):

$${{\rm{r}}}=\frac{{k}_{B}T}{h}\exp \left(\frac{-{E}_{a}}{{RT}}\right)$$
(2)

where R stands for the ideal gas constant and h signifies the reduced Planck constant. The r for the control and target films is calculated as 9.10 × 1010 s-1 and 1.63 × 109 s-1 at the operation temperature of 323 K70. The above results demonstrate a significant hindrance in ion migration within the target films, which is consistent with the aforementioned reduction in defect density and beneficial for mitigating hysteresis in PSCs, as well as enhancing device performance and photo-thermal stability. To directly observe the phase segregation phenomenon in perovskite films, time-dependent steady-state PL was measured. As depicted in Fig. 3h, i, after 60 min of irradiation, the PL peak position of the control shifts from 754 nm to 766 nm, accompanied by broadening of the peak width. In contrast, for the target, the PL peak shifts from 752 nm to 756 nm without a significant change in its width. These results collectively indicate that the target films exhibit improved morphology at the buried surface, alleviated interface strain, accelerated hole extraction dynamics, suppressed non-radiative recombination, and inhibited ion migration.

Device performance of single junctions and tandems

In order to evaluate the effect of MBC on device performance, the inverted PSCs with the structure of glass/ITO/NiOx/Me-4PACz/perovskite/PCBM/BCP/Ag was fabricated. The WBG perovskite Cs0.1FA0.65MA0.25Pb(I0.8Br0.2)3 (1.67 eV) precursor solution with the concentration of 1.25 M was used to deposit perovskite films with the thickness of 580 nm. The distribution of photovoltaic parameters in PSCs, prepared using different concentrations of MBC, is depicted in Supplementary Fig. 23 and Supplementary Table 3. As the concentration of MBC increases from 0 to 2.0 mg/mL, the photovoltaic performance initially exhibits an upward trend followed by a gradual decline, with the optimal concentration observed at 1.0 mg/mL. Compared to the control, the target devices with 1.0 mg/mL of MBC exhibited significantly improved average open-circuit voltage (VOC) from 1.23 ± 0.010 V to 1.26 ± 0.005 V, slightly increased short-circuit current density (JSC) from 21.25 ± 0.09 mA cm-2 to 21.50 ± 0.04 mA cm-2 and fill factor (FF) from 0.84 ± 0.01 to 0.85 ± 0.005, contributing to the average PCE increasing from 22.05 ± 0.28 to 23.33 ± 0.22. Figure 4a shows the current density-voltage (J-V) curves of the champion control and target devices. The control device shows the champion PCE of 22.33%, with the VOC of 1.24 V, JSC of 21.34 mA cm-2 and FF of 0.84. In contrast, the champion target device delivers a PCE as high as 23.56%, contributed by the VOC of 1.265 V, JSC of 21.54 mA cm-2 and FF of 0.86, which also exhibits a steady-state maximum power point (MPP) efficiency of 23.22% under the bias of 1.1 V (Supplementary Fig. 24a) and a slightly attenuated hysteresis index from 8.48% to 4.99% (Supplementary Fig. 24b and Supplementary Table 4). We also employed benzoic acid, benzimidazole and 4,5,6,7-Tetrahydro-1H-benzoimidazole-5-carboxylic acid (TBC) as additives to optimize device performance, whose molecular structures are depicted in Fig. 4b. Compared to MBC, benzoic acid and benzimidazole lack amphoteric properties, while TBC lacks a conjugated planar backbone. It is observed that all the additives led to a certain degree of reduction in device performance (Supplementary Fig. 25-27), thereby confirming the significance of our approach in constructing ACCMs and highlighting the importance of regulating the acid-base properties of these additives. Supplementary Fig. 28 shows the external quantum efficiency (EQE) to cross-check the JSC. The first derivative of the EQE curve verifies that the bandgap of the perovskite layer is ~1.67 eV, and the enhanced EQE within the absorption range results in an integrated JSC exceeding 21 mA cm-2, which aligns with the J-V findings. The long-term stability of the single-junction unsealed PSCs aged in ambient conditions (at 25 °C with a relative humidity of 20 ~ 30%, ISOS-D-1 standards) is demonstrated in Fig. 4c. The target displays an estimated T80 lifetime (the lifetime retaining 80% of the initial performance) of 4000 h, while the control shows the T80 lifetime of 1620 h. The enhanced shelf-life stability can be attributed to the suppressed permeation of humidity, which is supported by the increased water contact angle on the target films (Supplementary Fig. 29). Figure 4d poses the operational stability of PSCs, which is tracked at the MPP under continuous one-sun illumination (40 oC) according to the ISOS-L-1 testing protocols. The control displays a T80 lifetime of 400 h, while the T80 lifetime of the target is estimated to be 1300 h. The improved photostability is associated with the reduced density of defects and the enhanced crystallinity of the perovskite films.

Fig. 4: Perovskite single-junction and perovskite/silicon tandem device performance.
figure 4

a J-V curves of single-junction PSCs prepared with perovskite thin films. b Molecular structures of benzoic acid, benzimidazole and TBC. c Long-term stability of PSCs aged in ambient conditions. d Long-term continuous MPP tracking of PSCs. e Trap density in control and target PSCs. f FF limitation composed of nonradiative loss (gray area) and transport loss (red area). g Schematic stack of the monolithic perovskite/silicon TSCs. h J-V curves of single-junction PSCs prepared with 1 μm perovskite films. i J-V characteristics of champion perovskite/silicon TSCs. j Steady-state PCE for TSCs. k The long-term stability of TSCs aged under outdoor conditions (4 samples).

Full size image

In order to further investigate the underlying mechanism behind the performance improvement, we conducted a systematic examination of defect information, carrier recombination, and extractions from the perspective of intact devices. Supplementary Fig. 30 illustrates the capacitance-voltage (C-V) plots for a direct comparison of the built-in potential (Vbi). The target demonstrates a higher Vbi of 1.03 V in comparison to the control (0.91 V), indicating an enhanced separation and transfer of carriers within the target71,72,73. The thermal admittance spectroscopy in Fig. 4e is employed to investigate the distribution of trap states across different energy demarcations. The detail calculation method and relevant capacitance-frequency curves are provided in Supplementary Note 4 and Supplementary Fig. 31. The band of defects is categorized into three tiers. Band I is associated with a relatively shallow trap, whereas bands II and III are attributed to deeper-level defects. In this case, the target demonstrates a reduction in trap-state density within the energy range (Eω) values spanning from 0.35−0.50 eV74,75,76,77. Thus the density of both the shallow and deep defects are decreased, which is consistent with the prediction of DFT calculation and the results of XRD and PL. Light intensity (I)-dependent J-V measurements were conducted to further examine the charge carrier recombination categories in PSCs. The liner relationship between VOC and log-scaled I, as depicted in Supplementary Fig. 32, follows the Eq. (3)78:

$${V}_{{OC}}=\frac{n{k}_{B}T}{q}{ln}I$$
(3)

where n represents the ideal factor, and q signifies the elementary charge. The slope of the target (1.17 kBT/q) is lower than that of the control (1.30 kBT/q), indicating effective suppression of trap-assisted Shockley-Read-Hall recombination. The relationship between JSC and I can be described by the equation JSC ∞ Iα79,80,81, where the extracted α increases from 0.96 for the control to 0.99 for the target, manifesting a decrease in bimolecular recombination or a restriction on space-charge accumulation during the charge extraction process. The FF loss in PSCs is further analyzed by comparing the difference between the SQ limit (FFSQ) and the FF obtained from J-V analysis. The FF loss consists of two components: non-radiative losses and charge transport losses. The maximum FF (FFmax) without any charge transfer losses can be calculated as previously reported82,83. As shown in Fig. 4f, there is a significant reduction in non-radiative losses and a slight decrease in charge transfer losses in the target device. Therefore, the enhancement of FF in target PSCs primarily results from the inhibition of non-radiative recombination. Overall, the above results indicate that the addition of MBC improved the built-in electric field and reduced the defect density of various levels, thus facilitating the charge transfer process and suppressing the non-radiative recombination.

We then aim to apply this additive engineering to fabricate perovskite/silicon TSCs. Prior to that, we investigated the PCE of single-junction PSCs utilizing thick perovskite films, which is essential for achieving efficient TSCs. A high precursor concentration of 1.9 M was adopted to deposit perovskite films and the thickness increases to ~ 960 nm (Supplementary Fig. 33). Note that the inclined growth texture still persists in the thick film of target, while the control sample still exhibits voids at the buried interface. Supplementary Figs. 34–35 and Supplementary Table 5 illustrate the statistic photovoltaic parameters of the control and target PSCs, and the J-V curves of the champion devices are posed in Fig. 4h. Compared to the control, the champion target device demonstrates significantly higher PCE (22.41% versus 21.22%) and steady-state output power (22.27% versus 20.84%). It can be concluded that MBC exhibits a more pronounced capability in enhancing the quality of thick-film perovskite, thereby emphasizing the significance of studying PSCs employing thick perovskite films for TSCs. We then craft perovskite/silicon TSCs by utilizing the nano-textured silicon solar cells with the pyramid height of 360 nm. Figure 4i shows the cross-sectional SEM images of the intact TSCs, demonstrating that the perovskite layer can cover the pyramids completely. Supplementary Fig. 36 gives the distribution of photovoltaic parameters of TSCs and the J-V curves are plotted in Fig. 4i the champion PCE of the control and target TSCs are 29.46% and 31.57%, respectively. Supplementary Fig. 37 presents a third-party certified certificate with an efficiency of 30.24%. Compared to the control, the target devices show enhanced VOC (1.93 V versus 1.87 V) and FF (81.01% versus 78.58%), in addition to the comparable JSC (20.19 mA/cm2 versus 20.09 mA/cm2). Moreover, the hysteresis index of TSCs decreases from 6.92% to 2.5% (Supplementary Fig. 38 and Supplementary Table 6) and the target TSCs display a stable output power of 31.10% for 600 s at the bias of 1.58 V (Fig. 4j). The EQE spectra showed the integrated photo-current densities of 19.65 and 19.85 mA/cm2 for perovskite and silicon sub-cells, respectively, thus verifying a high matched current density (Supplementary Fig. 39). To evaluate the long-term stability of the TSCs, we tracked their MPP efficiency under continuous one-sun illumination at 40 °C following the ISOS-L-1 testing protocol. As shown in Supplementary Fig. 40, the unencapsulated target TSCs exhibit a T90 lifetime of ~750 h, whereas the control sample exhibits a 25% decline in initial performance after the same duration. In addition, we subjected the encapsulated TSCs to aging tests in an outdoor environment in Dalian, China (the weather is recorded in Supplementary Table 7). The control TSCs exhibited a T80 lifetime of 600 h (Supplementary Fig. 41). In contrast, the PCE of the target TSCs initially increased, followed by a gradual decline, and ultimately stabilized at their original performance level after 4000 h (Fig. 4k). Hence, MBC not only amplifies the device performance and long-term stability but also elevates the quality of the one-micron thick perovskite film, thereby yielding exceptional photovoltaic performance of the perovskite/silicon TSCs.

Discussion

In summary, we initiated this additive engineering based on benzoic acid and designed a functional ACCM for perovskites, named MBC, through the utilization of DFT calculations and Bronsted acid-base theory. Due to its strong interactions with perovskites, MBC demonstrates exceptional capabilities in modulating the crystallization process and passivating defects. The addition of MBC significantly retards the crystallization kinetics of perovskite films, leading to films with enhanced orientation and improved crystallinity. Consequently, it reduces the density of bulk defects and enhances film conductivity. Concurrently, MBC optimizes the morphology of buried interfaces while mitigating interface stress, thereby facilitating efficient charge transfer processes and suppressing non-radiative recombination. The implementation of MBC has enabled the fabrication of high-quality perovskite thin films, as well as micrometer-thick films, resulting in remarkable efficiencies of 23.55% and 22.47% for single-junction PSCs, respectively, which also demonstrate exceptional stability under various conditions. Finally, we achieved a PCE of 31.57% for a perovskite/silicon TSC based on MBC, while demonstrating exceptional stability under outdoor conditions. This study not only provides valuable insights into the design of effective additive molecules for perovskites but also presents a promising research direction to enhance the performance of TSCs.

Methods

Materials

N,N-dimethylformamide (DMF, 99.8%), dimethyl sulfoxide (DMSO, 99.9%) were purchased from Sigma-Aldrich. Methyl Acetate (MeOAc, 99%), isopropanol (IPA, 99.5%) were purchased from J&K. Lead iodide (PbI2, 99.999%), lead bromide (PbBr2, 99.99%), cesium iodide (CsI, 99.99%), methylammonium iodide (MAI, 99.99%), 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (BCP, 99.5%) were purchased from Xi’an Yuri Solar Co., Ltd. Formamidinium iodide (H2N = CHNH2I; FAI), [6,6]-phenyl-C61-butyric acid methyl ester (PCBM, 99.9%), and fullerene (C60, 99%) was purchased from Vizuchem Co., Ltd. Indium Tin Oxide (ITO coated glass, square resistance 15 Ω) was purchased from Advanced Election Technology Co., Ltd. [4-(3,6-dimethyl-9H-carbazol-9-yl)butyl]phosphonic acid (Me-4PACz, 99%), benzoic acid, benzimidazole and 2-methyl-3H-benzimidazole-5-carboxylic acid (MBC) were purchased from TCI. All salts and solvents were used as received without any further purification.

Preparation of nickel oxide films

The NiOx films were prepared via the electron-beam evaporation. The preparation detail and surface treatment of NiOx films are reported in the reference84.

Preparation of perovskite precursor

A 1.25 M Cs0.1FA0.65MA0.25Pb(I0.8Br0.2)3 wide bandgap perovskite precursor was prepared by adding 201.7 mg of PbI2, 68.9 mg of PbBr2, 16.3 mg of CsI, 69.9 mg of FAI and 24.9 mg of MAI in an alloyed solvents of DMF and DMSO with a volume ratio of 4:1. Then, the solutions were stirred overnight at room temperature in a N2 glovebox. The precursor solution is then filtered through a 0.45 mm polytetrafluoroethylene (PTFE) filter for later use.

Fabrication of single junction wide bandgap perovskite device

The ITO glass substrates were cleaned by using acetone and isopropanol in an ultrasonic bath for 30 min. Then, the substrates were dried by flowing nitrogen gas, followed by treatment in an ultraviolet-ozone chamber for 15 min. After that,[1] the substrates were transferred into an electron beam evaporation system. Before evaporation, the base pressure of the system was ~ 7 × 10-4 Pa. During the evaporation, a small flow of oxygen in the range of 15 sccm was present to improve the composition of the NiOx films, and the substrates were kept at room temperature in the whole process, the thickness of the NiOx was monitored by the crystal oscillator, and the thickness was set 26 nm. Then, the prepared substrate and the filtered perovskite precursor and other related solutions were transferred to the N2 glovebox for the next step. For using Me-4PACz as hole transportation layer, the solution of Me-4PACz in isopropanol with concentration of 0.5 mg/mL was deposited on the NiOx layer by spin coating at 3000 rpm for 30 s and annealed at 100 °C for 10 min on a hot plate. The isopropanol solution of alumina nanoparticles was spin-coated onto the substrate at 3000 rpm for 30 s. Then, 100 μL of perovskite solution was evenly dropped on ITO/NiOx/Me-4PACZ substrate and rotated at 1000 rpm for 10 s and 4000 rpm for 25 s, respectively. In the last 10 s of spin coating, 200 μL of MeOAc was dropped on the spinning substrates. Then, the perovskite films were immediately placed on a hot plate and annealed at 100 oC for 20 min. Then, PCBM (20 mg/ml in chlorobenzene) and BCP (saturated isopropanol solution) were spin-coated with spinning speeds of 4000 rpm for 30 s and 5000 rpm for 30 s, respectively. Lastly, 100 nm Ag was evaporated under high vacuum (<8 × 10-4 Pa) on the substrates to form electrodes.

Fabrication of silicon bottom solar cell

The silicon heterojunction bottom cell was fabricated on an n-type monocrystalline Czochralski silicon wafer with a thickness of ~150 μm. Both surfaces of the wafer were chemically textured in an alkaline solution to form randomly distributed pyramidal structures. Intrinsic and n-/p-doped amorphous silicon layers were subsequently deposited by plasma-enhanced chemical vapor deposition (PECVD). The rear electrode, consisting of ITO (70 nm) and Ag (300 nm), was prepared sequentially by RF sputtering and thermal evaporation. An ITO recombination layer with a thickness of about 20 nm was sputtered, with the active area defined as 1.44 cm² using a shadow mask.

Fabrication of tandem perovskite/silicon device

Silicon heterojunction (SHJ) bottom cells were fabricated on double-side textured wafers. The thicknesses of ITO recombination junction were set as 10 nm by sputtering. The 35 nm NiOx is then deposited on a silicon substrate using the same deposition method as a single-junction substrate. 1 mg/mL Me-4PACz in isopropanol was spin-coated at 3000 rpm for 30 s and thermally treated at 100 °C for 10 min. Then, the perovskite layer, whose deposition procedures were in consistence with the single-junction device deposition recipe, except for a 1.9 M of the solution, was employed. 30 nm of C60 was thermally evaporated as electron transport layer on top of the perovskite layer. A buffer layer of SnO2 (100 cycles) was then deposited by the atomic layer deposition (ALD) technique. Subsequently, 100 nm indium zinc oxide (IZO) top electrode with a sheet resistance of 60 Ω per square was deposited via sputtering at room temperature. Silver with thickness of 300 nm was thermal evaporated on the IZO and the p-side ITO through a shadow mask, respectively. The active area was calculated to 1 cm2. At last, 120 nm of MgF2 antireflection films were added through thermally evaporated at a rate of 2.5 Å/s to further enhance the light absorption.

Encapsulation of TSCs

The packaging steps of the devices is described as follows. First, silver-plated copper tape was used to connect the positive and negative terminals of the perovskite/silicon tandem solar cell. Then, a poly(ethylene-1-octene) (POE) film was laid flat at the center of the bottom glass substrate, the prepared device was placed on top, followed by another layer of POE. Butyl rubber was applied around the edges for sealing, and a second piece of glass was placed on top and pressed firmly. The assembled sample was then placed in a laminator and encapsulated at 130 °C for 10 min.

Device characterization

The J-V measurement was performed using a Keithley 2450 Source Meter under AM 1.5 G illumination from an Oriel 9600 solar simulator under air condition. A standard crystalline silicon cell was used to calibrate the light intensity of solar simulator before measurement. For single-junction PSCs, the devices were measured with a scan rate of 0.02 V/s and a delay time of 1 ms with the reverse scan range from 1.3 V to -0.1 V and the forward scan from -0.1 V to 1.3 V. Meanwhile, perovskite/silicon tandem devices were tested in the reverse scan direction of 2.0 V ~ -0.1 V and forward scan direction of -0.1 V ~ 2.0 V in 0.02 V steps. The external quantum efficiency (EQE) spectra were obtained by using a QE system (EnliTech), Mott-Schottky and electrochemical impedance spectroscopy (EIS) measurements were performed using an impedance spectroscopy analyzer (ZENNIUM, Zahner, Germany). SCLC measurements were obtained using a Keithley 2450 Source Meter under dark conditions, and devices were tested in forward scan direction 0 V to 3 V with step size of 0.005 V.

Films characterization

The morphology of the films was characterized by field emission SEM (SU8020). XRD patterns of various films were performed by using a Dandong DX-2700BH with Cu Kα1 irradiation (λ = 1.541 Å). UV-vis absorption spectra were recorded using a U-4150 Hitachi UV-vis spectrophotometer through transmittance. X-ray photoelectron spectroscopy measurements were carried out on an X-ray photoelectron spectrometer (Thermo ESCALAB 250Xi). Fourier transform infrared (FTIR) spectra was obtained using an Infrared spectrometer (Nicolet iS50, Thermo). Photoluminescence spectra (PL) and time-resolved photoluminescence (TRPL) were measured on a fluorescence spectrometer (Pico1000, Light-Stone CO, LTD) using the time-correlated single photon counting technique. The static water contact angles were measured by contact angle meter (DSA100). Time of-flight secondary ion mass spectrometry (ToF-SIMS) measurements were carried out using M6. For detection of negative and positive ions, Cs+ and O2- ions were used with ion energy of 2 keV, respectively. The AFM and the c-AFM measurement was performed by using an atomic force microscope (BRUKER ICON), where a SCM-PIT-V2 probe was used for the c-AFM detection according to the previous literature. Bare ITO glass substrates were coated with perovskite films with and without MBC. c-AFM current mapping was performed in contact mode with an applied bias between the perovskite film and the conductive probe. The conductive substrate is connected to the circuit to ensure the formation of a loop. Femtosecond transient absorption spectra (TA) were recorded by using an ultrafast TA spectrometer (Time-Tech Spectra, LLC). Femtosecond pump-probe TA measurements were performed at an appropriate power density (2.5 µJ cm-2). The pump pulse with a wavelength of 480 nm and a duration of 230 fs generated via a second harmonic generator (SHG) was used to excite all the samples, and the probe beam (from 619−848 nm) was detected by a high-speed spectrometer.

DFT calculations

All density functional theory (DFT) calculations were conducted using the Cambridge Serial Total Energy Package (CASTEP) with a plane-wave and pseudopotential approach. The physical properties of FAMAPb(IBr)3 were determined using the generalized gradient approximation (GGA) combined with the Perdew-Burke-Ernzerhof (PBE) functional. For these calculations, projector-augmented plane-wave pseudopotentials (PAPWP) were applied, while norm-conserving pseudopotentials were used specifically to assess the optical characteristics of the perovskite materials. A plane-wave basis set was used to solve the Kohn-Sham equations, aiming to minimize the total energy of the functionals.

To model the (100) surface structure, an optimized FAMAPb(IBr)3 unit cell was employed, incorporating a vacuum layer of ~30 Å along the z-axis. Structural optimization of the perovskite systems was performed with the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm, enabling the relaxation of both atomic positions and lattice parameters. After completing structural optimization, the materials’ optical and electrical properties were subsequently evaluated. Norm-conserving pseudopotentials incorporating Koelling-Harmon relativistic treatment were used to analyze the electrostatic interactions between valence electrons and the ionic core. Convergence criteria included limits of 2 × 10-5 eV/atom for total energy, 0.05 eV/Å for maximum force, 0.1 GPa for maximum stress, and 0.002 Å for maximum displacement, along with a self-consistent field (SCF) tolerance of 2 × 10-6 eV/atom and a cutoff energy of 700 eV.

Surface energy was calculated using the following Eq. (4):

$${Surface\; Energy}=\frac{1}{2\,A}\left({E}_{{suraface}}-{E}_{{Bulk}}\right)$$
(4)

Here, A represents the surface area, \({E}_{{suraface}}\) denotes the surface energy, and \({E}_{{Bulk}}\) indicates the bulk energy.

Reporting summary

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