Introduction

All-perovskite tandem solar cells1,2,3,4, combining wide and narrow bandgap subcells5,6,7,8, demonstrate the potential to surpass the Shockley-Queisser efficiency limit of single-junction photovoltaic devices9,10. While the efficiency of all-perovskite tandem solar cells is progressing rapidly, device stability becomes a critical issue currently hindering their commercial potential11,12. Photovoltaic devices under real-world operating conditions must endure operational temperature extremes to 85 °C. However, current high-efficiency Pb-Sn narrow bandgap perovskite solar cells (PSCs) are often doped with thermally unstable methylammonium (MA), whose propensity for thermal decomposition compromises device longevity13,14. Although replacing MA with thermally stable Cs enhances the perovskite thermal stability15,16,17,18, it often compromises film quality and reduces PCE17,19,20,21,22. This is mainly because low-solubility Cs-based23,24,25,26,27 perovskite tends to preferentially crystallize at the perovskite precursor solution film/anti-solvent (AS) solution interface during the AS-assisted crystallization process, forming a dense perovskite top-surface barrier layer (Supplementary Fig. 1). This continuous dense Cs-rich surface layer impedes the AS solvent extraction through the perovskite precursor solution, resulting in excessive solvent retention and asynchronous crystallization23,24,25,26,27. The retained excess solvent rapidly escapes during subsequent annealing processes, creating numerous voids at the perovskite buried interface. This ultimately induces an inhomogeneous Cs distribution and compromises the device performance.

Balancing the prior precipitation of dense perovskite at upper surface and the efficient solvent extraction during the AS-assisted crystallization process is critical for improving the photovoltaic(PV) performance of FACs-based Pb-Sn perovskites. The strong solvent extraction capability of AS hampers the effectiveness of conventional crystallization-delay strategies, such as adding dimethyl sulfoxide (DMSO)28 or Lewis acid to form intermediate-phase engineering29,30,31,32,33,34,35. Meanwhile, the trapped solvent, unable to be effectively removed, leads to the formation of voids and defects during thermal annealing.

Here, we deploy a surface tension modulation strategy using a strong multi-Lewis-base modulator to maintain sustained solvent-extraction channels (SSC) —a mechanism that prevents the formation of a continuous, dense thin perovskite layer at the perovskite/AS contact surface and preserves continuous solvent extraction channels throughout the AS process. Unlike conventional additive strategies that primarily regulate the intrinsic crystallization of perovskite36,37,38,39, SSC focus on maintaining efficient solvent-extraction channels during anti-solvent treatment, similar to solvent-extraction modulation concepts proposed in other material systems40. We reported a binary -COOH additive, L-Glutamic acid hydrochloride (Glu), added into perovskite precursor solutions to regulate the crystallization process of FACs-based Pb-Sn perovskites, aiming for high-performance and enhanced thermal stability of devices (Supplementary Fig. 2). The NH3+ group in Glu has a higher electrostatic potential and is likely to fill possible A-site vacancies(e.g., Cs⁺/FA⁺), while its dual -COOH groups tend to coordinate more strongly with Pb2+/Sn2+32,41,42. Such enhanced coordination effect enables Glu to modify the surface energy of the precursor solution, resulting in a higher heterogeneous nucleation barrier. As a consequence, the nucleation density is reduced, leading to fewer nucleation sites and the formation of larger perovskite grains. During the AS-assisted crystallization process, these larger grains pack less densely than smaller ones, exhibiting more channel-like structures between grains compared with the control perovskite surface. Therefore, Glu acts as the key additive that implements the SSC concept by maintaining sustained solvent-extraction channels during AS treatment.

This strategy enables a delicate balance between perovskite precipitation and efficient solvent extraction during the AS-assisted crystallization to achieve high-quality FACs-based Pb-Sn perovskite films. As a result, the champion MA-free mixed Pb-Sn perovskite solar cell yields the highest PCE of 22.7%. We further fabricated an all-perovskite tandem solar cell with a PCE of 29.2% (certified 29.2%), representing the highest certified PCE for MA-free all-perovskite tandem solar cells. The FACs-based mixed Pb-Sn perovskite solar cell with indium zinc oxide (IZO) electrode remained over 80% of its initial PCE after 800 h of thermal aging at 85 °C, demonstrating the state-of-the-art thermal stability for MA-free Pb-Sn perovskite devices. As a result, this approach achieves a combination of high power conversion efficiency and thermal stability in MA-free Pb-Sn perovskite solar cells.

Results

MA-free Pb-Sn perovskite with SSC

We used a one-step anti-solvent process to fabricate the MA-free Pb-Sn mixed perovskite films with a composition of FA0.85Cs0.15Pb0.5Sn0.5I3. However, due to the low solubility of Cs-based species, the preferential formation of a dense film at the perovskite/AS solution contact interface during the AS-assisted crystallization process inhibits efficient extraction of the internal solvents by AS. To address this issue, we implemented the SSC strategy by incorporating multi-Lewis base modulators into precursor solutions with different coordination ability (see “Methods”). We investigated three representative molecules with increasing numbers of carboxyl groups—glycine hydrochloride (Gly, one -COOH), Glu (two -COOH), sodium citrate (Sc, three -COOH) (Supplementary Fig. 3). Optical microscopy revealed anti-solvent (AS) extraction channels in the discontinuous 3D perovskite surfaces. The fraction of uncovered surface domains increased with the number of carboxyl groups in the additive, rising from 1.78% (control) to 5.17%, 7.11% and 15.72% for Gly, Glu and Sc, respectively (Fig. 1a and Supplementary Figs. 4 and 5).

Fig. 1: Mechanism of SSC-regulated crystallization process.
Fig. 1: Mechanism of SSC-regulated crystallization process.
Full size image

SSC denotes the perovskite film processed with 2% L-Glutamic acid hydrochloride (Glu) additive, which enables sustained solvent-extraction channels. a Optical microscopy images (100 × objective) of freshly AS-treated perovskite films, captured immediately after anti-solvent treatment and prior to annealing. The fraction of discontinuous domains increased from 1.78% (control) to 7.11% (SSC-Glu). b Contact angles of the corresponding perovskite precursor solution droplets on PEDOT:PSS. c 1H NMR spectra of pure Glu, pure FAI, FAI with Glu, and CsI with Glu additive dissolved in DMSO-d6. d Semi-in situ optical microscopy (50 × objective) tracking the morphological evolution of perovskite precursor films without (control) and with SSC over time. The images were acquired under an inert atmosphere without anti-solvent treatment or annealing. e XRD patterns of mixed Pb-Sn films processed from perovskite precursor solutions without and with SSC before the antisolvent treatment. f GC of the control and SSC treated wet perovskite films. g Schematic illustration of the SSC mechanism during AS-assisted crystallization process.

To better understand this process, we measured the viscosity of the precursor solution and the contact angles of the corresponding perovskite droplets on PEDOT:PSS coated substrates. The viscosity increased with the number of -COOH groups, due to enhanced coordination and steric hindrance (Supplementary Table 1). Correspondingly, the contact angle of the precursor solution on the substrate increased progressively with the number of -COOH groups (Fig. 1b and Supplementary Fig. 6). According to classical nucleation theory, the energy barrier for heterogeneous nucleation (\(\Delta {G}_{{Het}}^{*}\)) is proportional to the geometric factor f(θ), which is a monotonic function of the contact angle θ. A larger θ corresponds to a higher f(θ), and thus a higher \(\Delta {G}_{{Het}}^{*}\). This thermodynamic relationship implies that the higher nucleation energy barrier induced by SSC not only increases the critical radius, but also reduces the nucleation density, allowing fewer nuclei and ultimately leading to larger grain sizes (Supplementary Note 1). These theoretical insights align with our morphological observations: the control films with smaller contact angles undergo denser nucleation and yield smaller, more tightly packed grains. From a geometric standpoint, smaller particles tend to stack more densely, whereas larger ones leave more interstitial volume, contributing to increased film porosity. It is further demonstrated that the modulator additives are able to effectively adjust the superficial energy of the precursor solution, resulting in a higher heterogeneous nucleation barrier and a more restrained nucleation process, enabling more spatially distributed and uniform crystal growth. Scanning electron microscopy (SEM) images of the resulting perovskite films revealed not only enlarged grain sizes from control to Gly, and further to Glu, but also a marked improvement in grain uniformity (Supplementary Fig. 7). However, the Sc film did not follow this trend, which instead displayed poor surface crystallinity with non-uniform and small grains(Supplementary Figs. 8 and 9). These findings underscore the importance of balancing coordination strength for designing additive molecules. The Glu with two -COOH groups facilitates the formation of better perovskite films by regulating the crystallization process. Meanwhile, the excessively strong coordination—as observed in the case of Sc—may disrupt the controlled crystallization process and compromise film quality, which is consistent with previous reports that over-coordinating additives tend to induce non-uniform nucleation and hinder crystal growth37,41,43,44. To exclude the influence of amino groups and sodium salts, we also compared Formamidine hydrochloride (FACl) and NaCl as comparative additives (Supplementary Figs. 3 and 7–9). Both resulted in poor morphology, confirming that the improved film formation in Glu-treated samples primarily originates from balanced -COOH coordination rather than amino groups and sodium salts. Accordingly, Sc was excluded from further discussion in subsequent sections.

To investigate the crystallization modulation mechanism of SSC, we systematically explored its coordination chemistry and dynamic interactions with precursor components through a suite of chemical characterization techniques. FTIR spectra revealed that the characteristic C = O stretching peak of Glu shifted from 1699.78 to 1716.17 and 1668.93 cm−1, respectively, implying the strong coordination interaction between the -COOH groups of Glu and Pb²⁺/Sn²⁺ through Lewis acid-base adduct (Supplementary Fig. 10a). Compared to the commonly used monocarboxylic acid Gly33 with a single -COOH group, Glu with dual -COOH groups demonstrated a stronger effect (Supplementary Fig. 10b). The dicarboxylic modulator Glu induced more significant shifts in the -COOH stretching vibrations when mixed with PbI₂ or SnI₂. This stronger interaction likely arises from the dual -COOH groups in Glu, which can chelate metal centers more effectively than Gly. This strong coordination increases the effect of steric hindrance in the solution, thereby raising the viscosity and further supporting the observed increase in contact angle. Based on the comparative device performance of these modulators, Glu devices with higher average PCE were ultimately selected for the SSC strategy (Supplementary Fig. 11).

To complement these coordination analyses, we further employed ¹H nuclear magnetic resonance (NMR) spectra to investigate the effects of Glu incorporation. The ammonium proton resonance of Glu at 8.49 ppm shifted upfield upon mixing with FAI/CsI, while the dual resonances of FAI (8.65/8.99 ppm) shifted downfield (Fig. 1c), suggesting that Glu interacts and coordinates with the perovskite precursor components rather than new species formation and helps regulate the cation distribution45. X-ray diffraction (XRD) patterns of Glu-PbI₂ and Glu-SnI₂ mixtures exhibited distinct diffraction peaks absent in pure components (Supplementary Fig. 12), suggesting the intermediate phase formation via Glu-perovskite coordination. X-ray photoelectron spectroscopy (XPS) revealed a shift in the binding energy of the Sn signal on both the surface and bottom sides of the Glu film compared to the control (Supplementary Fig. 13), indicating a change in the Sn chemical environment. Concurrently, Glu reduced the proportion of Sn4+ peak area in Sn element peak area, suggesting suppressed Sn²⁺ oxidation and reduced defect density. Dynamic light scattering (DLS) further revealed that the addition of Glu increased the colloidal size of the perovskite precursor by 38% compared to the control (Supplementary Fig. 14), further supporting the presence of stronger coordination and a higher heterogeneous nucleation barrier46.

Based on a systematic comparison of film morphology, crystallinity, and device performance at different concentrations (Supplementary Figs. 15–17), a 2% concentration of Glu was identified as the optimal condition to achieving the SSC strategy, and was therefore used throughout this study. To directly probe the effect of SSC on perovskite crystallization kinetics, we performed in situ optical microscopy on spin-coated precursor films without applying antisolvent inside a nitrogen-filled glovebox. This setup enabled direct observation of the solvent evaporation-driven crystallization process. The nucleation and growth process of SSC perovskite crystallization can be significantly retarded compared to the control (Fig. 1d). These observations indicate that SSC not only delays nucleation but also significantly suppresses the crystal growth rate. X-ray diffraction spectra of wet films before AS treatment (Fig. 1e) further support this conclusion. The characteristic perovskite phase diffraction peaks appeared much earlier in the control film than in the SSC one, confirming a slower phase evolution. Collectively, these findings demonstrate that strong coordination between Glu and metal ions leads to a decelerated crystallization process, influencing final film morphology and microstructure. To further probe the crystallization process, we monitored the digital photos of the film during anti-solvent deposition (Supplementary Fig. 18). The control sample exhibited a darkening-lightening-darkening behavior, consistent with the transient formation and collapse of a dense surface layer. To delve deeper into the impact of SSC on perovskite film formation, we conducted in situ PL spectroscopy during both the spin-coating and annealing stages (Supplementary Fig. 19). The transient PL peak evolution of the control film (appearance-disappearance-reappearance) mirrors the changes in Supplementary Fig. 18, confirming the temporary formation of a dense surface layer that hinders solvent extraction. In contrast, the SSC film displayed a more continuous PL evolution without such interruptions, indicating more uniform crystallization and an unobstructed solvent-extraction pathway. Due to the PL detection fiber has an unavoidable absorption feature near ~ 940 nm that can produce an apparent two-peak appearance in the raw spectra (this effect and its manifestation in in-situ PL have been discussed previously47).

Gas chromatography (GC) quantification showed that SSC films retained < 15% residual DMF/DMSO versus > 32% in controls post-AS treatment (Fig. 1f), confirming enhanced solvent extraction capacity. The coordinated interplay between enlarged early-stage crystallites and sustained channel architecture enables complete removal of coordinated solvents, establishing a pathway for high-quality perovskite crystallization.

To elucidate the mechanism of SSC in suppressing surface preferential crystallization during the AS process, we proposed a schematic illustration comparing the control and SSC systems (Fig. 1g). In the control film (left), rapid surface crystallization forms a dense perovskite layer that obstructs efficient AS extraction. In contrast, the SSC system (right) balances the interfacial precipitation of dense perovskite and the AS deep solvent extraction during the AS-assisted crystallization through the coordination of Glu with perovskite precursor constituents. SSC can generate interstitial channels during AS-assisted crystallization. These channels prevent premature closure of extraction path ways by breaking compact perovskite crystallization, thereby preserving permeable networks for AS extraction process. This schematic illustration highlights that SSC does not solely delay crystallization but actively maintains sustained solvent-extraction during the AS stage—a key mechanistic distinction from conventional additive-assisted crystallization.

We investigated the effect of the SSC on the morphology of MA-free Pb-Sn alloyed perovskite films. Cross-sectional SEM images (Fig. 2a) and the corresponding buried interface SEM images (Supplementary Fig. 20) further revealed significant interfacial voids at the perovskite/substrate interface in the control film, indicating optimized crystallization kinetics enabled by the SSC strategy. To further distinguish our approach from conventional solvent-coordination control, we investigated the effect of DMSO content on film formation. Excessive DMSO leads to the formation of larger and more numerous voids at the perovskite interface, whereas the SSC strategy effectively alleviates this issue by preserving efficient solvent-extraction pathways during crystallization (Supplementary Figs. 21 and 22). To further examine the vertical distribution of Cs, we conducted cross-sectional EDS elemental mapping (Supplementary Figs. 23). The control film shows clear Cs accumulation near the surface and a gradual decrease toward the substrate, whereas the SSC film exhibits a much more uniform Cs distribution across the film thickness, supporting that the SSC treatment suppresses the formation of a Cs-rich surface layer. Time-of-flight secondary-ion mass spectrometry (TOF-SIMS) depth profiling unambiguously revealed the presence of a pronounced vertical composition gradient of Cs in control films (Fig. 2b and Supplementary Figs. 24 and 25), due to the preferentially surface crystallization of Cs-rich phases. Although the Cs⁺ profiles in Fig. 2b are plotted on a logarithmic scale for clarity, the corresponding linear-scale data (Supplementary Fig. 24) clearly show that the slope difference between the control and SSC films appears from the beginning of the measurement, confirming the existence of a steeper Cs gradient in the control film. Remarkably, SSC significantly flattens this gradient, homogenizing the vertical Cs distribution. This improvement can be attributed to the intrinsically low solubility of Cs-based salts, which often leads to premature surface crystallization during solution processing, particularly in FA-Cs mixed systems. By modulating the crystallization kinetics, SSC prevents the early aggregation of Cs-rich phases, thus enabling more uniform incorporation of Cs throughout the film. We further explored the carrier dynamics behavior of perovskite films to evaluate the impact of the SSC strategy on the film quality.

Fig. 2: Characterization of mixed Pb-Sn perovskite films with SSC treatment.
Fig. 2: Characterization of mixed Pb-Sn perovskite films with SSC treatment.
Full size image

a Cross-sectional SEM image of mixed Pb-Sn perovskite films without (control) and with SSC treatment. b TOF-SIMS depth profiles of Cs elements for control and SSC mixed Pb-Sn perovskite films, plotted on a logarithmic scale to show the elemental depth distribution. c Steady-state PL spectra of the control and SSC perovskite films deposited on ITO-PEDOT:PSS substrates. d time-resolved PL spectra of control and SSC Pb-Sn perovskite films deposited on bare glass (excitation from the glass sides). The decay curves were fitted with biexponential components to obtain fast and slow carrier lifetimes. e Electroluminescence quantum efficiency (EQEEL) versus current density for the control and SSC devices. f Estimated Quasi-Fermi level splitting (QFLS) of control and SSC perovskite.

The steady-state photoluminescence (PL) emission of the SSC film measured from both sides were identically stronger and higher than those of the control film, indicating suppressed non-radiative recombination due to reduced defect density (Fig. 2c). Notably, the emission peaks of the control film displayed a slight blue shift between its top and bottom surfaces, while the SSC sample showed identical emission peaks, reflecting improved vertical homogeneity. These results further support that the SSC strategy effectively mitigates the preferential surface crystallization of Cs-rich phases, leading to a more homogeneous Cs distribution throughout the film. We further performed time-resolved PL (TRPL) from both the perovskite side (Supplementary Fig. 26) and the glass side (Fig. 2d). The SSC film exhibited significantly longer carrier lifetimes than the control film, suggesting suppressed non-radiative recombination and reduced carrier capture loss. Considering the top-to-bottom crystallization pathway of perovskite films48, this observation is consistent with Glu gradually migrating toward the bottom during film growth, enabling effective defect passivation at the perovskite/substrate interface(Supplementary Fig. 27). Meanwhile, TOF-SIMS analysis (Supplementary Fig. 28) shows that a small amount of Cl- remains mainly at the buried interface. Together with the FACl comparison experiment, this indicates that Cl- plays a secondary passivation role rather than being the dominant factor in crystallization regulation. It is noteworthy that this spatial distribution is conducive to the defect passivation at the buried interface of perovskite films by Glu. We measured their UV-vis absorption spectra (Supplementary Fig. 29a) and then calculated their bandgap values (Supplementary Fig. 29b), which revealed negligible differences between control and SSC-treated films. Electroluminescence quantum yield (ELQY) was further used to analyze charge-carrier recombination49. At current densities equivalent to the short-circuit current density (JSC) under 1-sun illumination, the ELQYs of the control and SSC devices were 0.12% and 1.04%, respectively (Fig. 2e). To quantify the interfacial non-radiative recombination losses50,51, the photoluminescence quantum yield (PLQY) of perovskite was measured to estimate quasi-Fermi level splitting (QFLS) and voltage losses (Fig. 2f and Supplementary Table 2). The higher PLQY and ΔQFLS values of SSC films imply a significant suppression of non-radiative recombination channels in perovskite films, which is usually accompanied by an increase in VOC to enhance device performance. These results further validated that the reduction of defect density in Pb-Sn perovskite film was achieved after SSC treatment. Therefore, in addition to regulating the solvent-extraction pathways, Glu also behaves similarly to previously reported amino acid additives, which can modulate perovskite crystallization and passivate defect states, thereby enhancing both efficiency and stability of perovskite solar cells32,41,42,44,52. Based on the above results, the SSC-treated perovskite film can passivate the acceptor vacancy defect, which can reduce the probability of non-radiative recombination of carriers and improve the optoelectronic performance.

Photovoltaic (PV) performance of perovskite solar cells with SSC

We fabricated FACs-based mixed Pb-Sn PSCs without and with SSC treatment. The perovskite solar cells employed a device structure of ITO/PEDOT:PSS/Perovskite/C60/ALD-SnO2/Cu, where ITO is indium tin oxide, PEDOT:PSS is Poly (3,4-ethylenedioxythiophene)-poly (styrene sulfonate). Glu outperformed the control and other additives with a single -COOH group33,42, highlighting the superior effectiveness of the SSC strategy, which benefits from the stronger coordination achieved by dual -COOH functional groups (Fig. 3a and Supplementary Fig. 30). The SSC devices (aperture area of 0.049 cm2) delivered a champion PCE of 22.7% (stabilized at 22.5%), with an open-circuit voltage (VOC) of 0.871 V, a short-circuit current density (JSC) of 31.8 mA cm−2, and a fill factor (FF) of 81.9% (Fig. 3b and Supplementary Fig. 32). The increased JSC is ascribed to higher light absorption in the near-infrared range, as indicated by the external quantum efficiency (EQE) spectra (Supplementary Fig. 33). Furthermore, we also counted the histogram of PCE values of the SSC narrow-bandgap PSCs (Supplementary Fig. 34). The devices exhibited an average PCE of 21.76 ± 0.36% with a narrower distribution than the control devices.

Fig. 3: PV performance of perovskite solar cells.
Fig. 3: PV performance of perovskite solar cells.
Full size image

a Comparison of PV performance for control and SSC narrow-bandgap PSCs processed in the same runs. The box lines indicate the standard deviation, and the center represents the mean value. b The J-V curves of the champion control and SSC-treated device under reverse scan. c Cross-sectional SEM image of an all-perovskite tandem solar cell. d The J-V curve of the champion SSC tandem cell with an aperture area of 0.049 cm2. e EQE curves of the champion SSC tandem cell.

We fabricated monolithic all-perovskite tandem solar cells using the optimized narrow-bandgap (NBG) perovskite with a device aperture area of 0.049 cm2. The wide-bandgap (WBG) perovskite had a composition of FA0.8Cs0.2Pb(I0.6Br0.4)3. The tandem device consists of a ~ 450 nm WBG sub-cell and a ~ 900 nm NBG sub-cell, as shown in the cross-sectional SEM image (Fig. 3c). The WBG PSC showed a PCE of 20.1%, with a VOC of 1.348 V, a JSC of 17.8 mA cm−2, and a FF of 83.6% (Supplementary Fig. 35). Comparative analysis with control tandem devices showed improvement of all photovoltaic figures of merit for SSC tandem devices (Supplementary Fig. 36). The champion SSC tandem solar cell exhibited a high PCE of 29.2% (stabilized at 28.8%), with a VOC of 2.171 V, a JSC of 16.1 mA cm−2, and a FF of 83.72% (Fig. 3d, Supplementary Fig. 37 and Table 3). The histogram of PCE values of tandem solar cells showed a narrow distribution, with an average PCE of 28.37 ± 0.39% (Fig. 3d). The integrated JSC values of the front and back subcells from EQE spectra (Fig. 3e) were 16.1 and 16.1 mA cm−2, respectively. Furthermore, we sent one of the best-performing tandem solar cells to an accredited independent laboratory and achieved a certified PCE 29.2% (29.2%) at reverse (forward) scan (Supplementary Fig. 39). The tandem device delivered a certified stabilized PCE of 28.8% (Supplementary Fig. 39). This represents the highest certified PCE for MA-free all-perovskite tandem solar cells.

Thermal stability of MA-free Pb-Sn perovskite solar cells

In addition to the efficiency, long-term stability is another important parameter for evaluating the practical potential of the device. Therefore, we evaluated the thermal stability potential of MA-free Pb-Sn alloyed PSCs in real applications. We conducted ageing tests on unencapsulated mixed Pb-Sn PSCs by maintaining them on a hotplate at 85 °C in a N2-filled glove box (H2O, < 0.01 ppm; O2, < 0.01ppm). The unencapsulated SSC devices exhibited better thermal stability compared to the control devices, attributed to the improved perovskite film quality. In contrast, the FAMA devices exhibited a more rapid degradation immediately (Fig. 4a), underscoring the inferior thermal stability of FAMA absorbers under thermal stress.

Fig. 4: Thermal stability of MA-free Pb-Sn alloyed perovskite solar cells.
Fig. 4: Thermal stability of MA-free Pb-Sn alloyed perovskite solar cells.
Full size image

a Performance evolution of unencapsulated FACs-based (SSC and control) and FAMA-based devices aging at 85 °C under dark in N2 glovebox. c Schematic illustration of the original and modified structures of all-perovskite tandem solar cells. b, d Performance evolution of unencapsulated devices with different HTL (b) and different electrode (d) aging at 85 °C under dark in a N2 glovebox.

In addition to the perovskite absorber itself, charge transport materials also play a crucial role in affecting long-term stability and efficiency of PSCs. PEDOT:PSS is the most commonly used hole transport layer (HTL) for efficient mixed Pb-Sn PSCs, despite the fact that the deleterious interfacial reaction between the acidic PEDOT:PSS and Pb-Sn perovskites under thermal stress can lead to perovskite degradation and deteriorate the carrier extraction53. To further evaluate the impact of HTL selection, we proceeded to compare the thermal stability of mixed Pb-Sn PSCs fabricated with PEDOT:PSS and without any HTL (Fig. 4b). Although the device without HTL exhibited enhanced thermal stability, its PCE was markedly inferior to that of a device fabricated by PEDOT:PSS, indicating the urgent need for future research to develop alternative HTL materials that can balance both efficiency and stability.

Furthermore, the top electrode material has a pronounced impact on device durability. All devices discussed above employed Cu as the top metal electrode. To mitigate metal-induced degradation, we replaced the metal top contact with Indium Zinc Oxide (IZO) to eliminate its adverse impact on device stability54,55,56. We chose to capped the device with sputtered IZO in front of the metal electrode (Fig. 4c). This IZO interlayer effectively suppressed metal diffusion, thereby significantly enhancing the thermal stability of the devices.

It was demonstrated that the thermal stability of PSCs can be significantly enhanced by sputtered IZO. In addition to using a barrier layer, replacing Cu with inherently more stable electrode materials also contributed to improved device stability. For instance, substituting Cu with Au electrodes yielded perovskite tandem solar cells with enhanced thermal stability (Supplementary Note 2 and Supplementary Fig. 38). Thermal aging tests revealed that the device with only Cu electrode degraded to 60.6% of its initial PCE within just 180 h at 85 °C, whereas those employing IZO/Au electrodes remained over 80% of its initial PCE after 800 h, demonstrating state-of-the-art thermal stability for MA-free Pb-Sn perovskite devices (Fig. 4d). The preceding discussion of thermal stability results highlights the dual importance of developing both intrinsically stable mixed Pb-Sn perovskite and robust device structures to ensure the long-term operational stability.

Discussion

In summary, we developed a sustained solvent-extraction channel (SSC) strategy to regulate the crystallization of MA-free Pb-Sn alloyed perovskites, enabling the fabrication of high-performance and thermally stable NBG PSCs. This approach yielded a champion PCE of 22.7% for MA-free Pb-Sn alloyed PSCs and a PCE of 29.2% (certified 29.2%) for all-perovskite tandem solar cell, representing the highest certified PCE for MA-free all-perovskite TSCs. Moreover, unencapsulated FACs-based mixed Pb-Sn perovskite solar cell remained over 80% of its initial PCE after 800 h of aging at 85 °C, demonstrating state-of-the-art thermal stability for MA-free Pb-Sn perovskite devices. This strategy demonstrates the potential for simultaneous realization of high efficiency and stability in MA-free Pb-Sn perovskite solar cells (PSCs). This work establishes a foundation for the commercialization of this emerging PV technology by demonstrating devices with high efficiency and promising operational stability. Future work could explore integrating more stable interfacial materials under thermal stress, such as transparent conductive oxides and stable tunnel recombination layers. Concurrently, mechanistic studies of degradation pathways are essential to empower the rational design of targeted stabilization strategies for the long-term operational stability of all-perovskite tandem devices.

Methods

Materials

All materials were used as received without further purification. Lead halides (PbI2 and PbBr2, 99.99%), [2-(9 | H | -Carbazol-9-yl) ethyl] phosphonic Acid (99.9%), (2PACz, > 98%), and [2-(3,6-Dimethoxy-9H-carbazol-9-yl) ethyl] phosphonic Acid (MeO-2PACz, > 98%) were obtained from TCI Chemicals. Organic halide salts, including formamidinium iodide (FAI), cesium iodide (CsI), and phenylethylammonium iodide (PEAI, > 99%), were purchased from GreatCell Solar Materials (Australia). Poly(3,4-ethylenedioxythiophene) - poly (styrenesulfonate) (PEDOT: PSS) aqueous solution (Al 4083) was supplied by Heraeus Clevios (Germany). SnF2 (99%), Glycine hydrochloride (Gly, ≥ 99%), L-Glutamic acid hydrochloride (Glu, ≥ 99%), Anisole (99%), dimethylformamide (DMF, 99.8% anhydrous), and dimethyl sulfoxide (DMSO, 99.9% anhydrous) were purchased from Sigma-Aldrich. Sodium citrate(Sc, ≥ 98%) was purchased from Aladdin. Ethylenediammonium diiodide (EDADI, 99.5%) was purchased from Xi’an Polymer Light Technology (China). The C60 was purchased from Nano-C (USA).

Perovskite precursor solution

Narrow-bandgap (NBG) FA0.85Cs0.15Pb0.5Sn0.5I3 perovskite: A 2.0 M precursor solution was prepared by dissolving FAI, CsI, PbI2, and SnI2 in a mixed solvent of DMF and DMSO (volume ratio 3:1). The stoichiometric ratios of FAI:CsI and PbI2:SnI2 were fixed at 0.85:0.15 and 0.5:0.5, respectively, with a total A-site to B-site ratio of 1:1. SnF2 (10 mol% relative to SnI2) was added in the precursor solution. Additives (Gly, Glu, or Sc) were introduced at 1 mol% with respect to the total metal halide content. The precursor solution was stirred magnetically at room temperature for 2 h and filtered through a 0.22 μm PTFE membrane prior to use.

Wide-bandgap (WBG) FA0.8Cs0.2Pb(I0.6Br0.4)3 perovskite: The WBG precursor solution (1.25 M) was formulated in DMF/DMSO (4:1 by volume) using FAI, CsI, PbI2, and PbBr2, with molar ratios of 0.8:0.2 for FAI:CsI and 0.6:0.4 for PbI2:PbBr2. PEAI was added at 0.5 mol% relative to PbX2. The precursor solution was stirred at room temperature overnight, then filtered through a 0.22 μm PTFE membrane before making the perovskite films.

Mixed Pb-Sn narrow-bandgap perovskite solar cell fabrication

ITO-patterned glass substrates were cleaned using acetone and isopropanol and dried under a nitrogen flow. PEDOT: PSS was spin-coated on ITO substrates at 4000 r.p.m. for 30 s, followed by annealing at 150 °C for 10 min in ambient air. After cooling, the substrates were transferred into a nitrogen-filled glovebox for perovskite deposition. The NBG perovskite films were deposited using a two-step spin-coating process: an initial step at 1000 r.p.m. for 10 s with an acceleration of 200 r.p.m. s–1, followed by a second step at and 5000 r.p.m. for 80 s with a ramp-up of 1000 r.p.m. s–1. During the second spinning stage, anisole (200 μl) was introduced as an anti-solvent, which was swiftly dropped on the spinning substrate 20 s before the end of the second spin-coating step, using a pipette positioned approximately 3 cm above the film center. The anti-solvent treatment was performed under an inert N₂ atmosphere to ensure consistent solvent extraction and film formation. The substrates were then transferred to a hotplate and heated at 100 °C for 10 min. For surface passivation, EDADI post-treatment solution was dynamically spin-coated onto perovskite films at 4000 r.p.m. for 20 s and subsequently annealed at 100 °C for 1 min. After cooling down to room temperature, the substrates were transferred to a thermal evaporation system. Electron-transport layers were subsequently deposited by thermal evaporation of 20 nm C60 at a rate of 0.2 Å s−1. The ALD-SnO2 layer (~ 15 nm) was then deposited after C60 deposition at low temperatures (typically 70 °C). Finally, Cu electrodes (150 nm) were deposited by thermal evaporation at a rate of 1.0 Å s−1. The fabrication of FA0.85Cs0.15Pb0.5Sn0.5I3 perovskite solar cells is consistent with our previous works2.

For completeness, we note that EDADI post-treatment is applied as a standard surface passivation step in this work. A comparison of devices with and without EDADI (Supplementary Fig. 40) shows the passivation treatment leads to improved device performance, confirming its beneficial effect on defect passivation.

Monolithic all-perovskite tandem solar cell fabrication

Monolithic tandem devices were fabricated with the structure glass/ITO/NiO/SAM/WBG perovskite/C60/ALD SnO2/Au/PEDOT: PSS/NBG perovskite/C60/ ALD/Cu. Patterned ITO substrates were ultrasonically cleaned in acetone and isopropanol for 30 min each. NiO nanocrystals(25 mg ml−1 in deionized water) were synthesized following previous reported methods57 and spin-coated on ITO substrates at 3000 r.p.m. for 30 s, followed by annealing at 130 °C for 30 min in air. After transfer to a glovebox, SAMs (2PACz: MeO-2PACz = 75:25, 1 mM in isopropanol) were spin-coated on the NiO film at 6000 r.p.m. for 30 s, and then annealed at 150 °C for 10 min. Details on the Mixed SAMs for HTL has been discussed in our previous work58. The WBG perovskite layer was then deposited using a two-step spin program (2000 r.p.m. for 10 s, acceleration 200 r.p.m. s⁻¹; followed by 6000 r.p.m. for 40 s, acceleration 2000 r.p.m. s⁻¹), with anisole (200 μL) applied 20 s before the end of the second step. Films were annealed at 100 °C for 15 min. Subsequently, 20 nm C60 was deposited on top by thermal evaporation at a rate of 0.2 Å s–1, followed by deposition of a 20 nm SnO2 by the ALD system (Veeco Savannah S200) at low temperatures (typically 100 °C) using precursors of tetrakis (dimethylamino) tin(iv) (99.9999%, Nanjing Ai Mou Yuan Scientific Equipment) and deionized water. An ultrathin layer of Au clusters (~ 1 nm) was evaporated onto ALD-SnO2 prior to PEDOT: PSS deposition. PEDOT: PSS was spin-coated and annealed at 120 °C for 20 min in air. The NBG sub-cell was then fabricated on top using the same procedure as for single-junction devices, followed by deposition of 20 nm C60 film, ALD-SnO2 layer ( ~ 15 nm, 70 °C), and Cu electrodes (150 nm).

Characterization of solar cells

The J-V characteristics of single-junction solar cells were measured using a Keithley 2450 sourcemeter under the illumination of a solar simulator (EnliTech, Class AAA) at a light intensity of 100 mW cm−2 as checked with NREL calibrated reference solar cells (KG-5 and KG-0 reference cells were used for the measurements of the WBG and NBG solar cells, respectively). Measurements were performed in a nitrogen-filled glovebox at a scan rate of 100 mV s−1 with 10 mV voltage steps and a delay time of 100 ms. An aperture mask defined the active area (0.049 cm2). External Quantum Efficiency (EQE) spectra were recorded using a QE system (EnliTech) with monochromatic light focused on the device pixel and a chopper frequency of 20 Hz. The J-V characteristics of tandem solar cells were carried out under the illumination of a two-lamp high spectral match solar simulator (SAN-EI ELECTRIC, XHS-50S1), with spectral mismatch controlled within 100 ± 3% for each 50 nm interval in the wavelength range of 400–1000 nm. The illumination intensity of the solar simulator was adjusted to 100 mW cm−2 and verified using a calibrated crystalline silicon reference cell equipped with a quartz window (KG-0). EQE measurements were carried out in ambient conditions. During EQE acquisition, high-intensity LED bias lights with emission maxima at 850 nm and 460 nm were selectively applied to probe the front and rear sub-cells, respectively. No external electrical bias was applied throughout the EQE measurements of the tandem devices.

Thermal stability tests of solar cells

Unencapsulated devices were aged at 85 °C in a nitrogen-filled glovebox. J-V measurements were periodically recorded to evaluate performance evolution over time.

Steady-state and time-resolved photoluminescence(PL)

PL measurements were performed on a home-built wide-field microscope based on Olympus IX73. A diode laser with a wavelength of 485 nm served as the excitation light, and the laser power density incident on the film surface was maintained at 100 mW cm−2.The fluorescence of the samples was collected through a dry objective lens (Olympus LUCPlanFI 40×, Numerical Aperture (NA) = 0.6) and recorded using an Electron Multiplying Charge-Coupled Device (EMCCD) camera (iXon Ultra 888, Andor), after passing through an 850 nm long-pass filter (ET550LP, Chroma). Time-resolved photoluminescence (TRPL) measurements were carried out on a Horiba Fluorolog-3 time-correlated single photon counting system. The samples were excited from the glass side using a pulsed laser at 485 nm. The resulting PL decay curves were analyzed by fitting with a triple-exponential decay model, from which the fast and slow carrier lifetime components were extracted.

GC characterization

Pb-Sn perovskite films were deposited on glass substrates by spin coating and extracted by anti-solvent, and then put 20 those samples into 10 mL anisole to extract residual DMF and DMSO solvents. Then 0.1 mmol dodecane was added into the target mixed solvents as an internal standard. GC analyses were performed on a GC equipped with a flameionization detector and an Rtx@-65 (30 m × 0.32 mm ID × 0.25 μmdf) column. The temperature of SPL and FID is 280 °C. The injection volume of mixed solvents is 5.0 μL. The column temperature kept at 67 °C for 7 min and then rise to 250 °C at 20 °C /min retaining for 1 min. The signal peak of DMF, DMSO, dodecane appeared at 3.396 min, 3.984 min, 6.675 min, respectively.

Other characterizations

X-ray diffraction (XRD) measurements were performed on a Rigaku MiniFlex 600 diffractometer equipped with a NaI scintillation counter, employing monochromatized Cu Kα radiation with a wavelength of λ = 1.5406 Å. X-ray photoelectron spectroscopy (XPS) was conducted using the Thermo Scientific Al K-Alpha XPS system, with spectra recorded at energy steps of 0.1 eV. Ultraviolet-visible absorption spectra were acquired using a Lambda 950 UV-vis spectrophotometer. Scanning electron microscopy (SEM) imaging was carried out on a TESCAN microscope operated at an accelerating voltage of 2 kV.

Reporting summary

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