As a core development direction in third-generation photovoltaic (PV) technology, perovskite solar cells (PSCs) have garnered extensive attention in the field due to their unique photovoltaic properties and technical advantages^[1−3]. Its power conversion efficiency (PCE) has rapidly increased from 3.8% to 27%^{[4, 5]}, demonstrating immense potential for industrialization^{[6, 7]}. However, the instability of perovskite materials remains an unsolved problem^{[8, 9]}. This problem is often caused by the sensitivity of perovskite crystals to external environmental factors including moisture, oxygen, temperature, and light, which leads to their decomposition^[10−14]. Preparing higher-quality perovskite films has become a focus of attention. According to the LaMer mechanism and Ostwald ripening, the crystallinity and grain size of perovskites are closely related to controllable nucleation aggregation and growth rates^[15−17], which play a pivotal role in fabricating efficient and stable PSCs. Therefore, developing a strategy to modulate nucleation density and crystallization rates appears to be an effective approach for producing high-quality perovskite films.

Recently, several strategies have been developed to optimize the quality of perovskite films^[18−23], among which the introduction of polymer additives into perovskite precursor solutions is considered a key strategy for optimizing their crystallization properties. Due to their unique physical and chemical properties, polymers have drawn considerable research attention. On the one hand, their functional groups can interact with perovskites (e.g., coordination bonds or hydrogen bonds) to suppress defects and enhance crystallization performance^[24−26]. On the other hand, polymers also enhance the hydrophobicity of the films^[27−29], thus providing dual guarantees for high-performance PSCs. For example, Zheng et al. introduced polyacrylonitrile (PAN), which passivated the defects through hydrogen bond and coordination interaction, effectively addressing the crystallization and filling issues of FA-based perovskites in mesoscopic scaffolds, achieving a PCE of 18.33% in printable mesoscopic PSCs^[30]. Similarly, Zhang et al. investigated the relationship between polymer structure, crystallization and device performance. They found that the p-type quinone polymer PAQM-Bse actively induced perovskite nucleation and growth, proposing a "restricted assembly" growth mechanism, and successfully preparing high-quality perovskite films^[31]. Although polymer additives show great potential for enhancing the performance of perovskite devices, the introduction of unsuitable polymers may affect the charge transport properties of perovskite films, leading to device instability. Therefore, it is essential to choose more suitable polymer additives.

Here, we introduced a non-toxic and biodegradable polymeric material, poly-L-lysine hydrobromide (PLL), as an additive to obtain more efficient and stable PSCs by improving crystallization performance and locking perovskite crystals. The interaction between PLL and the perovskite is vital in creating premium-quality perovskite films. PLL modifies the aggregation state of colloids in the precursor solution, reduces nucleation density, and promotes the growth of larger perovskite crystals. Additionally, the widespread distribution of C=O and NH on the PLL interacts with the perovskite, effectively locking the perovskite crystals together, further enhancing film quality and structural stability. As a result, the introduction of PLL significantly improves device performance with an increased PCE from 22.20% to 23.66% and reduces hysteresis. Moreover, due to the PLL-induced structural reinforcement of the perovskite film and PLL's inherent hydrophobicity, the PLL-modified perovskite devices exhibit excellent long-term stability.

Lead iodide (PbI₂) and formamidine hydroiodide (FAI) were purchased from TCI and Youxuan Technology, respectively. Lead bromide (PbBr₂), N, N-Dimethylformamide (DMF) were sourced from Aladdin. Methylammonium bromide (MABr), cesium iodide (CsI) and Sprio-OMeTAD were obtained from Xi'an Polymer Light Technology Corp. Lithium bis ((trifluoromethyl)sulfonyl)azanide (Li-TFSI), methylamine hydrochloride (MACl), 4-tert-butylpyridine (t-BP) and dimethyl sulfoxide (DMSO), were gained from Innochem. Poly-L-lysine hydrobromide (PLL) was purchased from Adamas. Additional materials were purchased from Sinopharm or Alfa and were utilized directly and did not require purification or processing.

The structure of the device is ITO/SnO₂/(CsPbI₃)_0.025 (FAPbI₃)_0.9(MAPbBr)_0.075/Spiro-OMeTAD/Au. The manufacturing process is as follows：Firstly, the ITO glass was successively put into diluted detergent, deionized water, and anhydrous ethanol for ultrasonic cleaning for 20 min each. After that, it was dried at 60 °C to remove surface liquids. Once dried, UV ozone was carried out for 20 min to remove residues on the ITO surface. The electron transport layer (ETL) was prepared using an aqueous solution of SnO₂ with NH₄Cl at concentrations of 1 and 2 mg∙mL⁻¹. The preparation ratios were as follows: for each case, there was 1 mL of SnO₂ dispersion, 6 mL of ultrapure water, and 7 mg (for the 1 mg∙mL⁻¹ NH₄Cl solution) or 14 mg (for the 2 mg∙mL⁻¹ NH₄Cl solution) of NH₄Cl. The mixtures were stirred for 3 h. Then, 100 μL of the low-concentration (1 mg∙mL⁻¹ NH₄Cl) solution was taken and added dropwise onto the ITO surface at a spin-coating rate of 4000 rpm for 30 s. 100 μL of the high-concentration (2 mg∙mL⁻¹ NH₄Cl) solution was added dropwise at the 15th second. After annealing at 90 °C for 90 min in an air environment, the ETL was prepared. The perovskite precursor solution, which consisted of PbI₂, FAI, MABr, PbBr₂, MACl, and CsI, was dissolved into 1 mL of a DMF/DMSO mixture (precursor concentration of 1.4 M and DMF : DMSO = 4 : 1). 0.03 mg∙mL⁻¹ of Poly-L-lysine hydrobromide was incorporated with the perovskite precursor solution as an experimental group. The dissolution process was conducted at 60 °C with heating and stirring for 2 h. The perovskite layer was spin-coated in an air glove box with controlled humidity (RH ≈ 12%). 50 μL of precursor solution was taken and drop-cast onto the surface of the ETL at a spin-coating rate of 1100 rpm for 11 s, and then at 3500 rpm for 36 s. 150 μL of chlorobenzene (CB) was added dropwise as anti-solvent during the last 15 s. Then, the annealing process was carried out at 150 °C for 13−15 min to form the perovskite layer. 40 μL of the solution (Composition: 1 mL CB, 73.5 mg of Spiro-OMeTAD, 29 μL of t-BP, 17 μL of Li⁺ salt and 8 µL of Co³⁺ salt (Li⁺ salt and Co³⁺ salt concentrations of 500 and 400 mg∙mL⁻¹ acetonitrile solution respectively)) was taken and added dropwise onto the surface of the spin-coated perovskite layer at 4500 rpm for 30 s in a humidity-controlled air glove box to form the hole transport layer (HTL). Finally, device fabrication was finalized by depositing 80-nm-thick Au electrodes via thermal evaporation in a high-vacuum chamber (Quorum Q150T E Plus).

Infrared absorption spectra were measured by FTIR spectroscopy (Vertex80 + Hyperion2000). Samples were characterized using an in-situ X-ray diffractometer (D2 PHASER) with 2θ scanning from 5° to 50°. Dynamic light scattering (DLS) measurements (Malvern ZS90) provided particle size analysis, complemented by scanning electron microscope (SEM) imaging (GeminiSEM 300) for surface morphology. The absorption spectra of perovskite films were measured by a UV−vis spectrophotometers (U-3900). Steady-state photoluminescence (PL) spectroscopic measurements were measured using a fluorescence spectrometer (F-4700) at an excitation wavelength of 465 nm. Perovskite films surface topography were obtained using by ultra-high resolution scanning electron microscope (Regulus 8230). Measurements were made using a solar analog test system (Newport Oriel 94043A) and a digital in-situ meter (Keithley 2400) to obtain current−voltage characteristic curves of the perovskite devices. The fabricated devices featured an effective area of 0.09 cm⁻². Electrochemical impedance spectroscopy (EIS) was acquired in a dark environment using a CIMPS-pro system with the bias voltage set to −1 V, frequency range from 1 Hz to 1 MHz. Incident photon-to-current efficiency (IPCE) measurements were conducted on a Zennium CIMPS-pro system with spectral scanning from 300 to 900 nm. The water contact angles of perovskite films were measured using a fully automated contact angle surface energy tester (DSA25E). For the humidity aging experiments, the unsealed perovskite devices were placed ambient conditions (room temperature and relative humidity (RH) = 45 ± 5%). Additionally, heat aging (85 °C) was applied to unsealed perovskite films and devices.

Structural of PLL is depicted in Fig. 1(a). To verify the possible interaction between PLL and perovskite, we acquired FTIR measurements on pure PbI₂, pure PLL, and PbI₂ + PLL (Fig. 1(b)). The data show that the stretching vibrational peak of the C=O bond in the spectrum shifts from 1668.34 cm⁻¹ (for pure PLL) to 1667.37 cm⁻¹ (PbI₂ + PLL), while the bending vibrational peak of the Pb−I bond shifts from 1601.80 cm⁻¹ (pure PbI₂) to 1600.83 cm⁻¹. This change suggests that the C=O bond in PLL interacts with Pb²⁺, leading to the passivation of perovskite defects (such as uncoordinated Pb²⁺), which serves a critical function in enhancing the stability of perovskite films^{[32, 33]}. Fig. 1(c) illustrates the interaction force between PLL and perovskite. Based on this, we hypothesized that the long-chain PLL can be immobilized between the perovskite lattice to inhibit the migration of ions through the interaction (as demonstrated in Fig. 1(d)), to limit the formation of defect states and ultimately boost the crystallinity and device optoelectronic properties of perovskite films.

To examine the effect of PLL on perovskite crystals growth, the XRD patterns of the pristine and experimental (PLL) perovskite layers were presented in Fig. 2(a). Both samples display the characteristic peaks assigned to the (110) and (220) lattice planes of the perovskite phase at 14.1° and 28.2°, respectively, confirming that PLLs do not modify the perovskite crystal phase. Additionally, the (110) peak intensities and their corresponding FWHM values in Fig. 2(b) reveal that the PLL perovskite film exhibits significantly enhanced peak intensity and a smaller FWHM value, compared to the pristine sample. The increased peak intensity indicates enhanced crystallinity and preferential orientation, and the reduced FWHM, based on the Debye−Scherrer formula: D = Kλ/βcosθ (where K, λ, β, and θ represent the shape factor, X-ray wavelength, diffraction peak half width and Bragg diffraction angle, respectively) suggests larger grain sizes^[34]. These results collectively indicate that PLL positively enhances the crystallinity and preferential orientation and grain size of perovskite films.

As presented in Figs. 2(c) and 2(d), the optical properties of the perovskite films were further characterized using UV−vis absorption spectra and steady-state PL spectra. The light absorption properties of both pristine and PLL perovskite films were characterized by UV−vis spectroscopy. Enhanced absorption intensity observed in the PLL film correlates with improved crystalline quality and larger grain dimensions. This is further supported by PL measurements (Fig. 2(d)), where the PLL film demonstrates significantly higher peak intensities relative to the pristine sample. Generally, enhanced PL intensity implies decreased defect state concentration and suppressed non-radiative recombination in perovskite films. The PL spectra results reveal that the introduction of PLL effectively suppresses the formation of defect states in perovskite films.

To elucidate the role of PLL in modulating perovskite crystallization processes, we examined the distribution of colloidal particles in the precursor solution by DLS measurements, as shown in Figs. 3(a) and 3(d). The results indicate that pristine sample mainly presents aggregates with colloidal size less than 10 nm. In contrast, following the addition of PLL, the colloid size distribution was primarily characterized by two peaks at approximately 400 and 1280 nm. It can be seen that the introduction of PLL resulted in a significant increase in colloidal particle size, which may be attributed to the interaction between C=O in PLL and Pb²⁺ in the perovskite precursor solution. According to the classical nucleation theory, when the colloid particles are smaller than the critical size, the surface free energy dominates and leads to the dissolution of the particles. Whereas when the particle size exceeds the critical size, the bulk free energy dominates, allowing the large colloidal particles to grow directly as effective nuclei, thereby reducing the nucleation density and promoting the increase of crystal size^{[16, 35, 36]}. A large number of small-sized colloids (<10 nm) in the pristine system are prone to cause multi-directional growth and form randomly oriented small grains. In contrast, the larger aggregates (>400 nm) in PLL, due to their lower nucleation density, provide more sufficient growth time for the crystals, eventually forming perovskite films of larger size and higher crystallization quality. The surface morphology of perovskite films was characterized by SEM, verifying the DLS results, as shown in Figs. 3(b) and 3(e). The PLL samples exhibit a significantly improved crystal structure, with larger grains and fewer grain boundaries, compared to the pristine samples (the grain size statistics were shown in Figs. 3(c) and 3(f)). The smaller grain sizes in the pristine perovskite film led to a notable increase in the number of grain boundaries. It is notable that grain boundaries are defect-rich regions in perovskite materials, exacerbate non-radiative recombination and ion migration, thereby adversely affecting device performance^{[37, 38]}.

Furthermore, we characterized the performance of the PLL-based devices. To determine the optimal concentration of PLL, we evaluated the PV performance of the PLL devices with different concentrations (Fig. 4(a)), identifying 0.03 mg∙mL⁻¹ as the optimal concentration for achieving the highest PCE. Fig. 4(b) presents the J−V curves of the control and PLL devices under both reverse and forward scan directions (scan rate of 0.05 V∙s⁻¹). The pristine device exhibited parameters of V_oc = 1.17 V, FF = 78.20% and J_sc = 24.31 mA∙cm⁻². In contrast, owing to the higher crystal quality, reduced defect density and suppressed non-radiative recombination, the PLL device exhibited parameters of V_oc = 1.18 V, FF = 80.94% and J_sc = 24.71 mA∙cm⁻², resulting in a significant increase in PCE from 22.20% (for the pristine device) to 23.66%. In order to compare the hysteresis of the two devices, hysteresis factor was performed by applying the equation: HI = (PCE_reverse − PCE_forward)/PCE_reverse. The PLL PSC exhibits negligible hysteresis (0.70%) in comparison with the pristine device (2.65%). This enhancement can be ascribed to the suppression of ion migration as well as the reduction of defects. Detailed PV parameters are shown in Table 1. To verify the accuracy of J_sc, we performed the IPCE spectra on the devices. As shown in Fig. 4(c), the integrated J_sc values for the pristine and PLL PSCs are calculated to be 23.21 and 23.96 mA∙cm⁻², respectively, which match the results of the J−V curves. Due to its higher crystallinity, the PLL device exhibits a higher IPCE. To further explore the charge transfer of the devices, the EIS analysis was conducted on both the pristine and PLL devices (Fig. 4(d)). The results indicate that the PLL device displays significantly lower charge transfer resistance (R_tr) and notably higher recombination resistance (R_rec). This change in impedance characteristics suggests that the introduction of PLL not only optimizes charge transfer kinetics but also effectively suppresses the non-radiative recombination process of carriers.

Fig. 5 illustrates the distribution characteristics of PV parameters for the pristine and PLL PSCs through histograms and box plots. The PCE statistics in Fig. 5(a) show that the PLL devices exhibit more concentrated data distribution (with an average PCE of 23.66%), indicating their superior PV performance and higher process repeatability. For V_oc (Fig. 5(b)), the pristine devices display significantly lower average value and more dispersed distribution than the PLL perovskite devices. This can be attributed to the higher defect density in the pristine film, which exacerbates the non-radiative recombination process, shortens the photogenerated carrier lifetime and inhibits the formation of photogenerated potentials. Regarding the J_sc (Fig. 5(c)), the lower average J_sc value for the pristine group stems from the capture of carriers by the defect states, which act as charge traps, significantly reducing the effective collection efficiency of carriers. For the comparison of the FF parameter (Fig. 5(d)), the pristine devices also show a greater disadvantage due to various factors, such as poorer crystallization quality.

Finally, we investigated PLL's impact on perovskite film and device stability. Evaluation commenced with hydrophobicity measurements. Fig. 6(a) displays the water contact angle measurement results. The water contact angles of the pristine and PLL perovskite films are 62.45° and 73.45°, respectively, indicating that the PLL film exhibits stronger hydrophobicity. To evaluate the humidity lifetime of the devices, we performed humidity stability aging test on the pristine and PLL devices at room temperature with a humidity of 45 ± 5%, and we plotted the aging diagrams for different parameters (Figs. 6(b)−6(e)). The results show that the J_sc, V_oc, FF, and PCE of the pristine device drop to 84%, 88%, 82%, and 61% of the initial values after 4080 h, respectively. In contrast, the PLL devices maintain more than 96%, 98%, 97%, and 91% of the initial values for J_sc, V_oc, FF, and PCE after the same aging time. The large decrease in FF of the pristine device is caused by the poor crystal quality of the perovskite due to the higher defect density.

In addition, we evaluated pristine and PLL perovskite films and device thermal stability. We performed thermal stability analysis of the perovskite films by measuring the UV−vis absorption spectra (Figs. 7(a) and 7(b)) and changes of film morphology (Fig. 7(c)) at 85 °C. The results show that the pristine film degrades almost completely over time, while the PLL perovskite films show only fewer signs of decomposition. Similarly, the changes in the UV−vis absorption spectra indicate that after the same aging time, the pristine perovskite film shows a significant decrease in absorbance, while the PLL-based perovskite film can maintain more strong absorbance. Thermal stability evaluations were performed using the unsealed pristine and PLL devices aged at 85 °C in Figs. 7(d)−7(g). After 600 h, the PV performance of pristine devices exhibits significant degradation, with J_sc, V_oc, FF, and PCE dropping to 77%, 88%, 76%, and 52% of the initial values, respectively. In contrast, the PLL-treated devices demonstrate exceptional stability, retaining >97%, >98%, >95%, and >90% of their original J_sc, V_oc, FF, and PCE under identical conditions. This superior performance is attributed to the dual role of PLL in improving perovskite crystallinity and suppressing defect formation, thereby enhancing overall film and device stability.

In conclusion, we successfully introduced PLL into the perovskite precursor solution to substantially elevate the PV performance and stability of PSCs. By passivating defects and promoting the formation of larger crystals, PLL improves the film quality and reduces non-radiative recombination in perovskite films. Thereby, the perovskite device with PLL achieves a higher PCE of 23.66%, in comparison to 22.20% of the pristine device and significantly lower hysteresis. In addition, due to the hydrophobic long chain structure and the role in locked perovskite crystal structure of PLL, the PLL-based perovskite films and devices display the observably enhanced thermal and environmental stability. After aging at 45 ± 5% RH and 85 °C for 4080 and 600 h, the PLL-based perovskite devices still maintain 91% and 90% of the initial PCE, respectively.