Introduction

Perovskite/silicon tandem solar cells (TSCs) represent a promising pathway to overcome the efficiency limit of single-junction solar cells, with already demonstrated power conversion efficiencies (PCEs) exceeding 34%1. Most high-performance TSCs incorporate laboratory-scale silicon heterojunction (SHJ) bottom cells, including the world-record-efficiency perovskite/silicon TSC published by LONGi2. This achievement is primarily attributed to the high open-circuit voltage (VOC), strong near-infrared photoresponse, and low surface recombination losses of SHJ cells3,4,5,6. However, the relatively high manufacturing cost of SHJ cells restricts their widespread adoption in industrial applications. Alternative technologies such as passivated emitter and rear contact/tunnel oxide passivated contact (PERC/TOPCon) have gained significant traction in the industry due to their cost-effectiveness and potential for scaling to GW scale. To date, only a limited number of perovskite/silicon TSCs utilizing silicon bottom cells fabricated through this industrial technology route have been reported, with the highest published power conversion efficiency (PCE) for such devices reaching 31.3% (Fig. S1)7,8,9,10.

However, several critical challenges must be addressed to realize the full potential of perovskite/silicon TSCs, particularly in bridging the gap between current PCEs and the theoretical limit of 45%11. A major bottleneck lies in energy losses stemming from strong interfacial recombination at the perovskite/electron transport layer (ETL) interface12,13,14. C60, the commonly used ETL, contributes to non-radiative recombination due to interfacial defects and ionic migration, significantly suppressing the VOC and overall device performance15. Furthermore, operational stability remains a persistent issue for commercial deployment, as perovskite materials are susceptible to environmental and operational stresses. Traditional passivation strategies, such as metal fluoride16,17,18, ammonium salts19,20,21,22, or metal oxides23,24,25, often face trade-offs between passivation efficiency, ionic migration suppression, and long-term stability.

To address these challenges systematically, it is essential to establish a robust research approach that integrates theoretical modeling and experimental validation. This approach should aim to (i) identify and quantify energy losses at each interface within a tandem architecture, (ii) understand the fundamental mechanisms underlying these losses, and (iii) propose tailored solutions to mitigate them. Drift-diffusion simulations and density functional theory (DFT) calculations are powerful tools to analyze charge dynamics, interfacial interactions, and energy level alignment at the atomic scale, while experimental methods such as quasi-Fermi level splitting (QFLS) measurements, pseudo-J-V loss analysis, and X-ray photoelectron spectroscopy (XPS) provide practical validation of theoretical predictions. This combined approach would offer a systematic pathway to overcome the efficiency and stability bottlenecks in TSCs.

Based on this research approach, we propose a bilayer passivation strategy tailored to the specific challenges of perovskite/silicon TSCs. This strategy employs an ultrathin AlOx (~1 nm) layer deposited by atomic layer deposition (ALD) and a propane-1,3-diammonium iodide (PDAI2) layer between the perovskite absorber and C60. The ALD-AlOx layer provides conformal passivation of surface defects and inhibits ionic migration23,24, while the PDAI2 layer enhances the n-type doping, improving charge extraction and suppressing hysteresis20. By leveraging the complementary strengths of AlOx and PDAI2, the bilayer passivation simultaneously addresses energy loss and stability challenges, optimizing interfacial properties without compromising ionic transport dynamics. To demonstrate the efficacy of this approach, we fabricate monolithic perovskite/silicon TSCs with the proposed bilayer passivation. Systematic energy loss analysis reveals that the strategy significantly reduces non-radiative recombination at the perovskite/ETL interface, improves VOC, and enhances fill factor (FF). The resulting devices achieve a PCE of 31.6% (with a certified efficiency of 30.8%, aperture area of 1 cm2), one of the highest reported efficiencies for perovskite/silicon TSCs using industrial silicon bottom cell, alongside good operational stability, retaining 95% of their initial performance after 1000 h of maximum power point (MPP) tracking under 1-sun illumination (ISOS-L-1I). This study highlights the potential of our research approach to guide the design of high-performance, stable perovskite/silicon TSCs. It provides a framework for addressing similar challenges in the broader field of tandem photovoltaics.

Results

Systematic analysis of interfacial energy losses and limitations of mainstream tandem solar cells

The starting point of our analysis is a widely used architecture consisting of silicon bottom cell/NiOX/SAM/Perovskite/(passivation layer)/C60/SnO2/IZO/Ag (Fig.1a). Here, silicon bottom cells are Q.ANTUM-based industrial bottom Si solar cells, SAM refers to 4-(3,6-diphenyl-9H-carbazol-9-yl)butyl)phosphonic acid (Ph-4PACz)26,27, a self-assembled monolayer, and IZO represents zinc-doped indium oxide. Wide-bandgap perovskite, with a composition of (Cs0.05FA0.73MA0.22Pb(I0.77Br0.23)3 and an Eg of approximately 1.68 eV, is deposited with LiF passivation. As LiF is a well-established passivation strategy, it is employed here as the standard interface passivation layer16,17.

Fig. 1: LiF-based device performance loss analysis.
figure 1

a Device architecture of the perovskite/silicon tandem solar cell investigated in this study. b Absolute photoluminescence spectrum of a triple cation perovskite thin film (red dots, left y-axis) measured under equivalent one-sun conditions and Urbach energy (EU) obtained from photothermal deflection spectroscopy measurements (blue dots, right y-axis) of perovskite film on a quartz substrate. c QFLS values for quartz/PVSK, quartz/PVSK/LiF, quartz/PVSK/AlOx, quartz/PVSK/AlOx/PDAI2, and their corresponding counterparts with C60. For clarity, quartz is omitted from the sample names in the figure. d Summary of calculated FF losses, including non-radiative recombination loss and transport loss of devices. e Average PCE evolution of LiF-based perovskite/silicon tandem solar cells as a function of time under ISOS-D-2I protocol. The device is stored in a nitrogen-filled glove box at 85 °C in the dark for 1000 h. Average data are obtained from 3 cells and the error bars represent the standard deviation.

Full size image

In Fig. 1b, we first characterize our fabricated perovskite films on the quartz substrates with an absolute luminescence quantum yield system to determine their bandgap (~1.68 eV) and QFLS (1.276 V), as quartz is commonly regarded as a perfectly passivated surface with negligible recombination at the perovskite/quartz interface28. The QFLS values of the quartz/PVSK samples exhibit good reproducibility, as shown in Fig. S2. Based on detailed balance theory calculations, the radiative VOC for such a material is 1.384 V, indicating the non-radiative recombination centers in the pristine perovskite film cause a VOC loss of 108 mV. Furthermore, metal halide perovskites are known to be affected by an exponential Urbach tail in the absorption spectrum, which reflects the presence of localized states near the band edges29. Thus, to accurately evaluate the radiative limitation of our fabricated perovskite materials, we extract Urbach energy (24 meV) through photothermal deflection spectroscopy (PDS) by fitting the absorption edge, where the absorption coefficient exhibits an exponential dependence on photon energy. These two measurements confirm that the maximum VOC of the pure perovskite film is 1.32 V (see Supplementary Note 1). These findings highlight that optimizing the perovskite fabrication method and reducing non-radiative losses are critical challenges for unlocking the full potential of perovskite absorbers.

In addition, we aim to investigate the limiting factors responsible for the FF losses in device architectures. To achieve this, we perform intensity-dependent QFLS measurements (Fig. S3) of individual perovskite/transport layer combinations to assess how each interface contributes to the QFLS reduction observed in complete devices. First, to verify that phase segregation is not observed in our samples during the test period, we conduct time-dependent photoluminescence measurement under 532-nm continuous laser illumination (Fig. S4). During 60 min at 1-sun-equivalent illumination, our perovskite thin films exhibit no apparent low-energy peak and retains its PL spectral profile, suggesting that there is no significant phase segregation appearing. In individual perovskite/transport layer combinations, energy loss at the interface mainly stems from non-radiative recombination, which occurs at the SAM/perovskite interface, within the perovskite bulk, and at the perovskite/C60 interface. As illustrated in Fig. S5 and Fig. 1c, a systematic loss analysis based on pseudo-J-V curves is performed. The pseudo-J-V curves, derived from QFLS measurements under varying light intensities, confirm negligible series resistance losses. Compared to pristine perovskite (QFLS = 1.276 V), the SAM/PVSK (QFLS = 1.256) and PVSK/LiF (QFLS = 1.267) exhibited a slightly reduced QFLS. However, the deposition of C60 on PVSK/LiF caused a marked reduction in QFLS by 140 mV, resulting in a value of 1.127 V. Although this QFLS loss is smaller than the 209 mV drop observed in the PVSK/C60, the significant reduction suggests that interface loss primarily occurred at the perovskite and C60 interface, driven by the presence of C60.

AlOx, particularly when deposited as ultrathin layers via ALD, has emerged as a robust passivation strategy23,24,30,31. Al3+ ions can penetrate the perovskite bulk, interact with halide ions to suppress ionic migration and phase segregation, and simultaneously passivate defects at both the perovskite surface and grain boundaries. However, AlOx also acts as an efficient ion diffusion barrier, which can hinder the iodide-fullerene π-interaction. This interaction is moderately beneficial, as it contributes to the n-doping of C60, thereby enhancing charge transport and extraction while reducing hysteresis effects. Regarding this, PDAI2 is applied on top of AlOx, serving not only to chemically passivate the perovskite interface but also to facilitate n-doping19,20. Compared to the LiF-treated perovskite, PVSK/AlOx and PVSK/AlOx/PDAI2 exhibit similar QFLS values of 1.278 V and 1.276 V, respectively. In contrast, the QFLS loss upon C60 deposition is significantly lower for the AlOx/PDAI2-treated perovskite (18 mV), whereas the comparable loss for the AlOx-treated perovskite (137 mV).

Also, to understand whether the FF in device architectures is limited by insufficient charge transport or non-radiative recombination losses, we conduct an FF loss analysis on the aforementioned device architecture. The pseudo-fill factor (pFF) can be derived from pseudo-J-V curves. As an upper bound and reference to our measurements, the FF based on the detailed balance limit for a cell with a 1.68 eV bandgap is indicated (90.9%). Figure 1d summarizes the contributions to FF losses in pristine perovskite thin films and the corresponding thin films treated with LiF, AlOx, or AlOx/PDAI2. For pristine perovskite samples, the pFF value is 86.8%, 8.8% of FF loss is attributed to transport loss, and 4.1% to non-radiative recombination. Upon interface passivation, both transport loss and non-radiative recombination are reduced. The pFF values for the PVSK/LiF, PVSK/AlOx, and PVSK/AlOx/PDAI2 samples are 86.9%, 87.7%, and 87.0%, respectively. Specifically, the PVSK/AlOx/PDAI2 sample exhibits the lowest transport loss at 5.1%, followed by the PVSK/AlOx sample at 6.7%, and the PVSK/LiF sample at 6.9%. Regarding non-radiative recombination, the PVSK/AlOx sample shows the lowest loss at 3.2%, followed by the PVSK/AlOx/PDAI2 sample at 3.8% and the PVSK/LiF sample at 4.0%. Moreover, the stability of TSCs has been a significant challenge in commercializing perovskite/silicon TSCs. To investigate this, we execute a stability test on unencapsulated LiF-passivated perovskite/silicon TSC stored under nitrogen at 85 °C in the dark condition (ISOS-D-2I). The device retains only 74.5% of its initial PCE after 1000 h. Therefore, developing more effective strategies for recombination suppression, low resistance, and stable passivation is crucial to improve device performance and long-term stability.

Theoretical prediction of bilayer passivation strategy

The aforementioned systematic analysis of current perovskite/silicon TSC architecture highlights the central role of the interface between perovskite and ETL in energy losses and stability of TSCs. While fullerenes, such as C60, remain necessary due to their favorable properties, including high electron affinity, mobility, and efficient vertical transport, addressing interfacial limitations is essential for further performance improvements32,33,34. Also, passivation layers should act as a diffusion barrier, effectively suppressing the outward migration of species from the bulk perovskite (e.g., I−, CH3NH3−) and inhibiting the inward diffusion of external species (e.g., Ag+, H2O). To address these challenges, passivation strategies are needed to achieve defect passivation, energy level modulation, low diffusion affinity, and enhanced charge extraction, as illustrated in Fig. 2a. Additionally, to overcome the inefficiencies of traditional trial-and-error approaches, a research approach is crucial for accelerating progress in this field.

Fig. 2: DFT calculation and electronic properties of perovskite with or without surface treatment.
figure 2

a Schematic of bilayer passivation strategy. Differential charge density maps for AlOx on b PbI-terminated perovskite and c FAI-terminated perovskite, where electron depletion is indicated by green and electron accumulation in yellow. Projected density of states (PDOS) from density functional theory calculations for perovskite with different terminations: d FAI-termination; e PbI-termination. The control represents untreated perovskite, while AlOx denotes AlOx-treated perovskite. f Energy-level alignment derived from UPS for control, PDAI2-treated, AlOx-treated, and AlOx/PDAI2-treated perovskite samples. Here, EV refers to the valence band maximum, EC to the conduction band minimum, EF is the Fermi energy level, and EVAC represents the vacuum energy level. g Energy level diagram at the interface between AlOx/PDAI2-treated perovskite and C60. h Band alignment simulated via drift-diffusion modeling for solar cells with AlOx/PDAI2 treatment at the open-circuit condition. Here, Fn and Fp represent electron and hole quasi-Fermi level, respectively. i Simulated energy level profiles and charge carrier densities for AlOx/PDAI2 solar cell at the open-circuit condition.

Full size image

To design such an approach and develop effective passivation strategies, it is essential to understand the underlying interactions between the passivation material and the perovskite surface. In this context, we employ DFT calculations to explore the passivation effect of AlOx due to its feature of upscaling processing35,36,37. Initially, we optimize the perovskite architecture (Fig. S6), considering two distinct terminations: PbI termination and formamidinium iodide (FAI) termination. Differential charge density calculations (Fig. 2b, c) reveal significant charge transfer between the oxygen atoms in AlOx and the surface atoms of perovskite slabs. Furthermore, a comparison of the projected density of states (PDOS) before and after AlOx treatment demonstrates that, for the PbI-terminated perovskite, trap state density near the valence band edge is reduced by AlOx (Fig. 2e). Similarly, AlOx significantly reduces the trap density near the conduction band edge in FAI-terminated perovskites (Fig. 2d).

The electronic structure of the perovskite films is characterized using ultraviolet photoelectron spectroscopy (UPS, Fig. 2f). Compared to pristine perovskite with a work function (WF) of 4.36 eV, PDAI2-treated perovskite exhibits a reduced WF of 4.01 eV. Additionally, the Fermi level (EF) of PDAI2-treated samples shift upwards towards the conduction band minimum (EC), increasing the EF and valence band maximum difference (EF –EV) from 1.03 eV for pristine perovskites to 1.59 eV. This upshift of EF indicates typical n-type doping behavior for the perovskite materials and enhances electron transport. Notably, AlOx-treated perovskite shows the lowest WF (3.73 eV), while the AlOx/PDAI2 bilayer-treated sample exhibits a slightly increased WF of 3.84 eV. Meanwhile, the EF –EV value is 1.42 eV for the AlOx-treated sample and 1.56 eV for the AlOx/PDAI2-treated sample. The energy level diagram for the perovskite/C60 interface after AlOx/PDAI2 bilayer treatment is summarized in Fig. 2g. After bilayer passivation, the EF of perovskite shifts closer to the EC, resulting in a downward band bending at the perovskite surface. Simultaneously, n-type doping of C60 due to the migration of I− from PDAI238,39 drives the EF of C60 nearer to its lowest unoccupied molecular orbital, inducing similar downward band bending at the C60 surface. Consequently, the bilayer passivation strategy optimizes the energy level alignment between the perovskite and C60 and improves charge extraction efficiency.

Drift-diffusion simulations are employed to analyze band alignment and charge carrier density profiles at the perovskite/C60 interface at a device level, providing critical insights into the impact of interface treatments on charge density profiles, recombination behavior, and overall device performance. SCAPS-1D (a Solar Cell Capacitance Simulator) is employed to build the device model with the p-i-n architecture40. Detailed parameters are illustrated in Table S1. Drift-diffusion simulations under open-circuit conditions (Fig. 2h and Fig. S7) reveal a downward shift in the EF relative to the EV at the C60 interface, with the AlOx/PDAI2-treated perovskite exhibiting the most significant shift. We further analyze the charge density profiles of simulated devices with different surface treatments to figure out the carrier concentrations of devices at the open-circuit condition (Fig. 2i and S7; the simulation parameters are listed in Table S1). Owing to the efficient electron extraction by C60, the electron density in C60 is high while the hole density near the interface remains low, indicating reduced recombination processes. Among the devices with different interface treatments, the electron density in C60 shows minimal variation, but the hole density ranks as AlOx/PDAI2 < AlOx < PDAI2. The simulated J–V curves confirm the improved device performance (mainly from VOC) of different interfacial treatments (Fig. S9). This finding highlights the efficacy of AlOx/PDAI2 bilayer passivation in mitigating charge recombination, which is expected to improve device performance.

Experimental verification of bilayer passivation strategy

Previous theoretical results provide the detailed potential influence of interface treatments at the perovskite/C60 interface, highlighting the significant impact of AlOx/PDAI2 bilayer passivation in mitigating charge recombination. To validate these theoretical findings and assess their practical applicability, we conduct experimental analyses using XPS. The XPS results confirm the successful modification of AlOx and PDAI2 on the perovskite surface, demonstrating effective surface passivation and significant interactions between the interface materials and perovskite, which is evidenced by shifts in binding energies and the disappearance of metallic lead signals. As shown in Fig. 3a, the presence of AlOx in the AlOx- and AlOx/PDAI2-treated films is confirmed by the Al 2p peak at ~74.8 eV36. In the N 1 s orbital (Fig. 3b), two distinct peaks are observed, corresponding to the C = N bond of formamidinium (FA) at ~400.4 eV and the C–N bond of methylammonium (MA) or the PDA ligand at ~402.3 eV41. Compared to pristine perovskite films and AlOx-treated films, the C–N/C = N ratio increases significantly in PDAI2- and AlOx/PDAI2-treated films, indicating the successful incorporation of PDA ligands on the surface of the perovskite thin films. In pristine perovskite films, the Pb 4f7/2 and Pb 4f5/2 peaks appear at approximately 138.3 eV and 143.2 eV, respectively (Fig. 3c). Additionally, weak metallic lead (Pb0) signals are observed at ~136.5 eV and ~141.5 eV. However, after PDAI2, AlOx, or AlOx/PDAI2 surface modifications, the Pb0 signals completely disappear, demonstrating the effective passivation of surface metallic lead defects by all surface modifications42. Furthermore, the Pb 4f7/2 and Pb 4f5/2 peaks shift to lower binding energies in the modified perovskite films, indicating significant interactions between the interface materials and the perovskite. Among these modifications, the largest binding energy shift is observed in AlOx/PDAI2-treated films (~0.6 eV), followed by AlOx-treated films (~0.5 eV) and PDAI2-treated films (~0.2 eV). This trend is corroborated by the binding energy shifts of the I 3 d peaks (Fig. S10).

Fig. 3: Interface interaction, photoluminescence imaging, and morphological analysis.
figure 3

XPS spectra of the a Al 2p, b N 1 s, and c Pb 4 f core levels for the different perovskite films. The Pb0 signal is also marked in the c. d−k PL imaging of control, PDAI2-, AlOx-, and AlOx/PDAI2-treated films, with and without C60, on a silicon/HTL substrate. i-o AFM images of different perovskite films (top panels), with corresponding schematic diagrams (bottom panels). The size of AFM images is 0.5 × 1 μm. The schematic diagram depicts the morphological features extracted from AFM images: PDAI2 forms a homogeneous, undetectable thin film on the perovskite surface (yellow transparent layer); ALD-AlOx is uniformly distributed across the perovskite grain surfaces, with island-like structures emerging at the grain boundaries (blue dots in the grain boundaries); the density of island-like AlOx distribution at the perovskite grain boundaries is significantly reduced with AlOx/PDAI2 treatment.

Full size image

To assess the homogeneity of non-radiative recombination losses at the perovskite/ETL interface, photoluminescence (PL) images of different perovskite thin films, with and without C60 coverage, are obtained (Fig. 3d–k)43. Prior to C60 deposition, perovskite samples with different surface treatments exhibit comparable PL intensities, with the AlOx-treated sample showing slightly higher intensity, followed by pristine perovskite, AlOx/PDAI2-treated sample, and PDAI2-treated sample. Notably, the surface of the perovskite treated by AlOx exhibits non-uniformity, which is attributed to the heterogeneous distribution of AlOx across the perovskite grain surfaces and along the grain boundaries. However, subsequent deposition of the PDAI2 atop the AlOx-treated perovskite surface effectively renders the non-radiative recombination losses more uniformly distributed across the perovskite surface. Upon C60 deposition, the PL intensities of all samples decrease. The AlOx/PDAI2-treated perovskite shows the strongest PL intensity, followed by the AlOx-treated sample, then the PDAI2-treated sample, with pristine perovskite showing the lowest PL intensity. These results together indicate a high defect density at the perovskite/C60 interface, leading to significant non-radiative recombination losses. Importantly, the AlOx/PDAI2 treatment effectively reduces the defect density and optimizes the band alignment at this interface, thereby substantially suppressing non-radiative recombination and improving interface quality.

Figure 3i–o shows atomic force microscopy (AFM) images and corresponding schematic diagrams of perovskite film morphology to illustrate the features of different surface modifications. After spin-coating the PDAI2 on the perovskite surface (Fig. 3m), the morphology remains nearly identical to that of pristine perovskite thin films (Fig. 3l). Upon depositing an ~1 nm-thick AlOx layer on the perovskite surface using the ALD method, a conformal AlOx layer forms on the perovskite grain surfaces. While this part of AlOx is nearly undetectable in AFM images, nanoscale AlOx islands (bright spots) are clearly observed at the grain boundaries (Fig. 3n). For AlOx/PDAI2-treated perovskite films, similar bright spots are observed at the grain boundaries. (Fig. 3o). To validate the presence of AlOx islands at perovskite grain boundaries, we conduct high-resolution AFM imaging (1 × 1 μm) on AlOx/PDAI2 bilayer-treated perovskite thin films with varying AlOx thicknesses (0.5 nm, 1 nm, 1.5 nm, and 20 nm), see Fig. S11. For better comparison, pristine perovskite, PDAI2-treated, and AlOx-treated (1 nm) samples are also shown in Fig. S11. The thicknesses of the AlOx thin films are nominal values indicated by the ALD system. Height images clearly show AlOx islands along the grain boundaries, while the AlOx layer on the perovskite grain surface is nearly undetectable in ultra-thin AlOx-treated perovskite films, unlike pristine and PDAI2-treated samples. In AlOx/PDAI2-treated perovskite films, increasing AlOx thickness reduces the density of island-like AlOx at grain boundaries. At 20 nm thickness, these islands become nearly undetectable, indicating a transition from discontinuous island structures to a more continuous and uniform AlOx film. Figure S11h–n presents the corresponding AFM phase images, where regions of differing contrast indicate microstructures with distinct mechanical properties, such as friction, elastic modulus, composition, and viscoelasticity44. The pristine perovskite film and the PDAI2-treated perovskite film exhibit relatively continuous surfaces, resulting in minimal phase contraction. In contrast, the deposition of an ultra-thin AlOx layer leads to a significant increase in phase shift difference, which is attributed to the formation of a relatively non-uniform film on the perovskite surface. The island-like AlOx structures at the grain boundaries expose underlying perovskite regions, and the mechanical mismatch between AlOx and the perovskite contributes to the observed contrast. With increasing thickness of the AlOx layer, the surface becomes more uniform and the AlOx layer eventually fully covers the perovskite surface, leading to a reduction in phase shift variation. Top-view SEM images (Fig. S12) further confirm that a dense AlOx film covers the perovskite surface, along with nanoscale channels at the grain boundaries due to island-like AlOx.

The formation of these island-like AlOx can be attributed to an inhibited initial growth mechanism45, which results from the lack of ALD-reactive moieties on the substrate surface. During the initial stages of the ALD-based AlOx, precursor molecules nucleate at isolated sites on the perovskite surface. Subsequently, AlOx islands begin to grow across the surface, gradually expanding and eventually coalescing into a continuous film37,46. Subsequently, spin-coating a PDAI2 layer onto the AlOx-coated surface causes partial removal of the AlOx layer at grain boundaries due to the flushing effect of the solvent, resulting in a less dense AlOx coverage in grain boundaries. To confirm that the PDAI2 treatment does not completely remove the underlying AlOx layer, we use IPA to wash the surface of AlOx/PDAI2-modified perovskite films by spin-coating 10 times. The surface morphology before and after washing is analysed using SEM, along with EDX elemental mapping (Fig. S13). The SEM images show no significant morphological differences before and after IPA washing. Additionally, EDX mapping confirms the continued presence of Al and O elements associated with AlOx on the perovskite surface, further supporting the existence of the AlOx layer after PDAI2 deposition. The remaining gaps in the less dense AlOx layer at the grain boundaries provide nanoscale channels for direct contacts between the upper PDAI2 layer and the perovskite. This bilayer structure not only allows the PDAI2 layer to passivate areas of the perovskite that are not fully covered by the AlOx layer, but also mitigates excessive I− migration from perovskite bulk to C60, which occurs with only PDAI2-treated films. These synergistic effects may contribute to enhanced device performance and operational stability.

We perform light intensity-dependent QFLS measurements to systematically quantify the efficiency potential of the individual perovskite/transport layer combinations of the top cells in TSCs (Fig. 4a, Fig. S14, and Table S2). The results reveal that bilayer passivation induces slightly increased bulk and HTL interface losses. Specifically, the pristine perovskite film exhibits a bulk loss of 108 mV and an HTL interface loss of 4 mV, whereas the AlOx/PDAI2-treated film demonstrates a bulk loss of 112 mV and an HTL interface loss of 5 mV. After the deposition of the ETL, the bilayer treatment significantly reduces the VOC loss at the perovskite/ETL interface from 125 mV to 9 mV. Moreover, the bilayer passivation decreases the mismatch between the full-stack samples used for QFLS measurements and the device VOC, from 3 mV in the control sample to 1 mV. In addition to mitigating VOC loss, the FF loss due to transport resistance is also mitigated, decreasing from 4.2% in the pristine perovskite film to 2.1% in the bilayer-treated device. These results further demonstrate that our proposed bilayer passivation strategy could suppress non-radiative recombination processes and enhance carrier extraction efficiency.

Fig. 4: Photovoltaic characteristics and stability.
figure 4

a Loss analysis for FF and VOC of control and bilayer-treated devices. FF and VOC losses are extracted from pseudo-J-V measurements detailed in Fig. S14 and Table S2. b Reverse and forward J−V scans under 100 mW cm−2 AM 1.5 G standard solar irradiation. c The stabilized PCE outputs of the AlOx/PDAI2-treated cells. d Evolution of PCE of unencapsulated tandem solar cells with aging time at 85 °C in the nitrogen-filled glovebox. Error bars represent the standard deviation, calculated from 3 cells. e Continuous maximum power point tracking for the unencapsulated tandem solar cells under AM 1.5 G illumination in the nitrogen-filled glovebox. Calculated local electrostatic potential (ESP) for f AlOx- and g PDAI2-passivated perovskite thin film. Planar macroscopic average of ESP of h PVSK/AlOx and i PVSK/PDAI2 calculated as functions of fractional coordinate.

Full size image

Photovoltaic performance of tandem solar cells

Subsequently, monolithic perovskite/silicon TSCs are fabricated using the AlOx/PDAI2-treated perovskite light-absorbing layer in a device architecture comprising silicon bottom cell/NiOx/Ph-4PACz/perovskite/AlOx/PDAI2/C60/SnO2/IZO/Ag stack. The silicon bottom cells utilized in this work are fabricated using Q Cells’ Q.ANTUM technology. Ph-4PACz is used due to its good wettability and improved perovskite phase homogeneity27. The performance comparison of perovskite/silicon TSCs employing Ph-4PACz, 2PACz, and Me-4PACz is presented in Fig. S15. To improve the light management, MgF2 is applied as an anti-reflection coating. PDAI2- and AlOx-treated tandem devices are included as references. J–V curves are measured under simulated AM1.5 G illumination at an intensity of 100 mW cm−2. To optimize the efficiency of devices with bilayer passivation, tandem devices are fabricated with varying thicknesses and concentrations of AlOx and PDAI2. As shown in Fig. S16 a combination employing 1 nm of AlOx in combination with 0.3 mg·mL−1 of PDAI2 yields the most effective interface passivation. Fig. 4b presents the representative J–V curves, and averaged photovoltaic parameters, including VOC, short-circuit current density (JSC), FF, and PCE, summarized in Table S3. The AlOx/PDAI2-based tandem solar cell demonstrates a PCE of up to 31.6%, with a JSC of 19.91 mA cm−2, a VOC of 1.96 V, and a FF of 81.0% under reverse scan. In comparison, the PDAI2-based TSC achieves a PCE of 30.2%, with a JSC of 19.89 mA cm−2, VOC of 1.92 V, and a FF of 79.2% in the same condition, while the AlOx-based TSC exhibits a PCE of 29.3%, with a JSC of 19.90 mA cm−2, VOC of 1.85 V, and a FF of 79.7%. Notably, all devices display negligible hysteresis. The stabilized PCE output affords 31.3% for the AlOx/PDAI2-based cell (Fig. 4c). An unencapsulated AlOx/PDAI2-based device is sent to a third party for certification, where it achieved a certified PCE of 30.8%, a JSC of 19.77 mA cm−2, a VOC of 1.97 V, and a FF of 79.0% in the reverse scan, corroborating our internal measurements (Fig. S17). Statistical data from devices, as presented in Fig. S18, confirm the good reproducibility of our results and suggest that the improvement in PCE is primarily attributed to enhancements in VOC and FF. Fig. S19 shows the external quantum efficiency (EQE) spectra and optical reflectance of PDAI2-, AlOx-, and AlOx/PDAI2-treated TSCs integrated with the standard AM1.5 G solar emission spectrum. The AlOx/PDAI2-based tandem cell achieves integrated JSC values from EQE measurements of 19.89 and 19.73 mA cm−2 for the perovskite and silicon subcells, respectively. Similarly, the AlOx-treated tandem cell exhibits integrated JSC values of 19.76 and 19.55 mA cm−2, while the PDAI2-treated device records integrated JSC values of 19.82 and 19.63 mA cm−2 for the perovskite and silicon subcells, respectively. The reflectance spectra of the PDAI2-, AlOx-, and AlOx/PDAI2-treated devices are nearly identical, indicating that the differences observed in JSC are more likely attributed to variations in carrier transport or collection rather than optical effects. The EQE measurements reveal a slight deviation of approximately 0.8% compared to the JSC values obtained from J–V scans. We note that the EQE spectra of the silicon subcell are not shown for compliance reasons as shown in the attached nature photovoltaic report table.

Next, to evaluate the thermal stability of the TSCs, we conduct accelerated degradation testing following the ISOS-D-2I protocol. Unencapsulated devices are placed on a hotplate maintained at 85 °C in a nitrogen-filled glovebox. Periodically, the TSCs are removed and subjected to J–V characterization under AM1.5 G illumination in ambient conditions. The evolution of photovoltaic parameters is shown in Fig. 4d and Fig. S20. After 1000 h, the PCE of AlOx/PDAI2-based TSC retains 92% of its initial efficiency, with VOC decreasing by 1.1%, FF by 5.4%, and JSC by 2.0%. In contrast, the AlOx-based TSC retains 87% of its initial PCE, while the PDAI2-based TSC retains only 82% of its initial efficiency. Additionally, we subject unencapsulated tandem devices to operational stability testing under continuous 1-sun illumination (ISOS-L-1I; Fig. 4e). The spectrum of LED light source can be found in Fig. S21. The device treated with PDAI2 exhibits a rapid decline in PCE under maximum power point tracking (PCEMPP) during the first 200 h, falling below 38% of its initial value. In contrast, AlOx-treated device shows significantly improved stability, retaining 90% of their initial efficiency after 1000 h. The AlOx/PDAI2-treated device demonstrates the best operational stability, with only 5% degradation in PCE over 1000 h.

To elucidate the role of the AlOx/PDAI2 layer in enhancing the operational stability of TSC systems, we employ DFT calculations to determine the local potential distribution of the perovskite/AlOx layer (Fig. 4f) and perovskite/PDAI2 (Fig. 4g). AlOx exhibits an electron rich characteristic, which reduces the local electrostatic potential and establishes a larger potential difference between AlOx and the perovskite layer (Fig. 4h). The potential barrier formed by AlOx effectively suppresses ion migration outward from the perovskite. In contrast to AlOx layer, the PDAI2 molecules anchored to the surface cannot form either a blocking layer or an obvious potential barrier (Fig. 4i), unable to contain the ion migration. Phase images extracted from AFM results (Fig. S11) and SEM images (Fig. S12) confirm the presence of an AlOx layer on the surface of the AlOx/PDAI2-treated perovskite, which plays a crucial role in suppressing ion migration and thereby significantly enhances the long-term operational stability of the device. Fig. S22 presents a schematic illustration of the mechanism by which the AlOx/PDAI2 bilayer interface suppresses ion migration.

Discussion

In summary, we establish a research approach that systematically analyzes energy losses and guides the design of effective passivation strategies for perovskite/silicon TSCs. Using this approach, we propose a bilayer passivation strategy that sequentially combines ALD-grown AlOx and solution-processed PDAI2 on the surface of wide-bandgap perovskites. This strategy achieves precise interface modulation, addressing critical challenges in energy loss and operational stability. The bilayer-treated perovskites exhibit optimized energy level alignment and significantly reduced defect densities. DFT calculations of the electron localization function reveal that the ALD-deposited AlOx establishes a substantial potential difference within the perovskite, effectively serving as a barrier to ion migration and enhancing the stability of the perovskite interface. Concurrently, AFM observations indicate that the ALD process forms island-like structures at the grain boundaries of the perovskite, creating nanoscale localized contact regions. These localized regions provide a contact pathway between the second passivation layer PDAI2, and perovskite, enabling n-type doping by PDAI2 and facilitating efficient charge transport and extraction. This AlOx/PDAI2 bilayer passivation strategy effectively balances the trade-offs associated with iodide migration, significantly reducing non-radiative recombination losses at the perovskite/C60 interface.

Consequently, monolithic perovskite/silicon TSCs incorporating the bilayer-treated perovskites achieve a certified PCE of 30.8% and a representative PCE of 31.6% on a 1 cm2 aperture area—among the highest efficiencies reported for tandem devices based on industrial silicon bottom cells to date. Moreover, the bilayer-treated devices exhibit good thermal and operational stability compared to devices treated with either AlOx or PDAI2 alone, maintaining 92% of their initial efficiency after 1000 h of dark storage at 85 °C and retaining 95% of their performance after 1000 h of MPPT under 1-sun illumination.

Furthermore, future development of the AlOx/PDAI2 bilayer strategy may offer deeper insights and enable further performance enhancements. For instance, cross-sectional KPFM2 could understand the interfacial electric field and morphological distribution, while quasi–Fermi-level splitting mapping17 could assess the uniformity of the passivation layer. Additional experimental efforts—such as fine-tuning the ALD process (e.g., varying thickness, pulse time, or temperature), exploring alternative low-temperature ALD chemistries, or modifying the PDAI2 molecular structure—could further improve passivation quality and interface stability. These directions offer exciting opportunities for future research.

In conclusion, this study highlights the importance of a systematic research paradigm in addressing interfacial challenges and demonstrates the potential of bilayer passivation strategies for precisely regulating perovskite surface properties. The AlOx/PDAI2 bilayer demonstrates a pathway for achieving high-efficiency and durable perovskite/silicon tandem devices, advancing the commercialization of this next-generation photovoltaic technology.

Methods

Materials

(4-(3,6-Diphenyl-9H-carbazol-9-yl)butyl)phosphonic acid (Ph-4PACz, >99%, Luminescence Technology), lead iodide (PbI2, 99.99%, TCI), lead bromide (PbBr2, >98.0%, TCI), formamidinium iodide (FAI, >99.99%, Greatcell solar materials), methylammonium bromide (MABr, 99.99%, Dyenamo), cesium iodide (CsI, 99.9%, Alfa Aesar), propane-1,3-diammonium iodide (PDAI2, >99.5%, Luminescence Technology), fullerene-C60 (C60, 99.5%, Sigma-Aldrich), 2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (BCP, >99.5%, Luminescence Technology), magnesium fluoride (MgF2, ≥99.99%, Sigma-Aldrich), lithium fluoride (LiF, >99.99%, Luminescence Technology). All solvents including N,N-dimethylformamide (DMF, 99.8%), dimethyl sulfoxide (DMSO, ≥99.9%), 2-propanol (IPA, 99.5%), methanol (>99.5%) were ordered from Sigma-Aldrich. All materials were used as received without further purification.

Solar cells fabrication

Single-junction perovskite solar cells

The p-i-n type perovskite solar cells with the architecture ITO glass/NiOX/Ph-4PACz/perovskite/passivation layer/C60/BCP/Ag were fabricated as follows. ITO glasses (16 × 16 mm, sheet resistance 15 Ω cm−2, Luminescence Technology) were progressively cleaned by sonication with detergent, deionized water, acetone and IPA for 15 min each. The washed ITO glasses were dried by N2 flow. Post-cleaning, the ITO glasses underwent UV-ozone treatment for 10 min before being transferred to a N2-filled glovebox for film fabrication. Ph-4PACz (0.5 mg mL−1 in methanol) was statically spin-coated onto the cleaned ITO glass at 3000 rpm for 30 s, followed by annealing at 100 °C for 10 min. Then, 80 μL perovskite precursor solution was spin-coated at 1000 rpm for 10 s and 5000 rpm for 30 s onto the Ph-4PACz covered ITO substrate, 150 μL ethyl acetate as antisolvent was dripped on the films at 13 s before the end of the last procedure and then annealed at 100 °C for 20 min. The perovskite precursor solution (1.4 M) was prepared by mixing FAI, MABr, CsI, PbI2, and PbBr2 in DMF/DMSO mixed solvent (v/v: 4/1) with chemical formula Cs0.05FA0.73MA0.22Pb(I0.77Br0.23)3 + 3% PbI2. After the perovskite, a passivation layer was deposited, which can be LiF, AlOx, PDAI2, or AlOx/PDAI2. 1 nm LiF layer was deposited by thermal evaporation; c.a. 1 nm AlOx layer was deposited by the thermal atomic layer deposition (ALD) technique. The substrate temperature was maintained at 90 °C during ALD deposition and trimethylaluminum (TMA) precursor source and H2O source were both without heating. The pulse and purge time for TMA is 0.2 and 8.0 s with a 30 sccm N2, for H2O is 0.2 and 8.0 s with 30 sccm N2. 8 cycles were used; PDAI2 treatment was done by spin-coating. 0.3 mg mL−1 PDAI2 solution in IPA/CB mixed solution (v/v 1/1) was dynamically spin-coated at 4500 rpm for 25 s, and then annealed at 100 °C for 5 min. Then, 15 nm C60, 5 nm BCP, and 100 nm Ag electrodes were sequentially evaporated under a high vacuum (< 4 × 10−6 torr). A 100-nm MgF2 layer was thermally evaporated onto the back of the devices for the anti-reflection coating.

Perovskite/silicon tandem solar cells

Before deposition, the silicon bottom cells (Qcells, 25 mm × 25 mm) were washed with acetone and IPA in a spincoater process. The silicon bottom cells were then subjected to UV-Ozone treatment for 5 minutes before NiOx modification. A 15-nm NiOX film was sputtered from a NiOX target using 100 W power with pure Ar at 1 mTorr on the substrate. Then, the same SAM, perovskite (1.5 M), passivation layer, C60 deposition as described above was conducted on the Si/NiOx substrate. A 20 nm SnO2 layer deposited by ALD was used as buffer layer. The substrate temperature was maintained at 90 oC during ALD deposition with Tetrakis(dimethylamino)tin(IV) (TDMASn) precursor source at 70 oC and H2O source at room temperature. The pulse and purge time for TDMASn is 1 and 10.0 s with a 90 sccm N2, for H2O is 0.2 and 15.0 s with 90 sccm N2. 200 cycles were used. Subsequently, 45 nm IZO was sputtered from a IZO target through a shadow mask, using 190 W power with pure Ar and O2 at 1 mTorr. Ag finger with a thickness of 600 nm was thermally evaporated using a high-precision shadow mask. The finger width is approximately 75 μm. 100 nm MgF2 was eventually thermal evaporated on top of the Ag as an anti-reflection coating.

Solar cells characterization

The J−V characteristics of single-junction perovskite solar cells are performed by Keithley 2400 in a N2-filled glovebox at room temperature under AM 1.5 G illumination (100 mW cm−2) from a class AAA xenon-lamp solar simulator (Newport Oriel Sol3A). The solar simulator irradiation intensity was calibrated with a filtered KG5 silicon solar cell (Fraunhofer ISE CalLab). The J−V curves were obtained both in reverse (1.3 V to −0.1 V) and forward scan (−0.1 V to 1.3 V) with step size of 10 mV. For perovskite/silicon tandem solar cells, J−V measurements of were carried out in the air under LED-based solar simulator (WaveLabs Sinus 70) at room temperature. The solar simulator irradiation intensity was calibrated with a certified silicon solar cell (Fraunhofer ISE CalLab). The active area was defined by a black metal mask featuring an aperture with precisely measured area of 1.0 cm2. The devices underwent test through both reverse scans (2.1 V to −0.1 V, incrementing in 20 mV steps) and forward scans (−0.1 V to 2.1 V, with the same incremental step), conducted at a scan rate of 10 mV s−1. Delay time is 10 ms. The EQE was conducted with a PVE300 photovoltaic QE system (Bentham EQE system) in a nitrogen-filled glove box. Spectra in the wavelength range of 300 to 1250 nm for perovskite/silicon tandem solar cells were acquired using a chopping frequency in the range of 560–590 Hz and an integration time of 1000 ms. Due to the insufficient intensity of our LED light source, two bias LEDs were used for each subcell in the tandem devices to ensure that the subcell being measured was the current-limiting one. When measuring perovskite top cell, the tandem devices were light-biased by two IR LEDs with 780 nm and 940 nm peak emissions to saturate the silicon bottom cell. The silicon bottom cell is measured by saturating the perovskite top cell with a blue LED (465 nm) and a white LED. For MPP tracking of tandems, the unencapsulated devices were operated under 1 Sun LED illumination (WaveLabs Sinus 220). To evaluate the thermal stability of the tandem device, the unencapsulated devices were subjected to accelerated aging on a hot plate maintained at 85 °C inside a nitrogen-filled glove box. At regular intervals, the devices were removed for J-V characterization under ambient conditions and subsequently returned to the hot plate for continued thermal aging.

Perovskite film characterization

Pristine perovskite and perovskite covered with PDAI2, AlOx, and AlOx/PDAI2 passivation layers on ITO glass/Ph-4PACz substrates were investigated. X-ray photoelectron spectroscopy (XPS) was conducted using a K-Alpha instrument (Thermo Scientific), equipped with a monochromatic Al Kα X-ray Omicron XM1000 X-ray source (hʋ = 1486.6 eV), referencing the binding energy scale to the C 1 s signal. Ultraviolet photoelectron spectroscopy (UPS) was performed with an ESCALAB XI+ instrument (Thermo Fisher), utilizing a He(1) source (21.22 eV) under a negative bias of 5.0 V. Atomic force microscopy (AFM) measurements were executed with a Nano Wizard II microscope (JPK Instruments). PLQY measurements were carried out using a LuQY Pro setup (QYB). The samples were mounted inside an integrating sphere and a green laser (λ = 532 nm) was directed into the sphere via a small entrance port. The quasi-Fermi level splitting (QFLS) were measured under a series of light intensities using the same tool. PL-imaging was performed using a home-made setup. Two 467 nm LED bars, aligned opposite at 45° to the image plane, illuminate the sample to provide a homogenous excitation equivalent to 0.2 suns. The resulting photoluminescence is imaged by a CMOS camera equipped with a macro zoom lens and a 695 nm absorptive long-pass filter. The resulting images were measured with an exposure time of 100 ms and a background correction to exclude stray light and camera noise.

Photothermal deflection spectroscopy

The samples were mounted inside a quartz cuvette filled with a thermo-optic liquid (3 M Fluorinert FC-72). The excitation source consisted of a halogen lamp coupled to a 250-mm focal length grating monochromator, providing tunable light beam wavelengths for spectral scans, additionally modulated at 10 Hz with a mechanical chopper. The PDS experiments were performed in the transverse configuration, with a probe laser beam (670 nm) passed close and parallel to the sample surface in the area excited with the pump beam. The probe beam deflection induced by heat transfer was detected with a quadrant silicon photodiode and measured synchronously using a lock-in amplifier (Stanford Research Systems SR830). Urbach energy was derived from the absorption edge where the absorption (A) is exponentially related to the photo energy via:

$${\rm{A}}\left(E\right)={{\rm{\alpha }}}_{0}\exp \left(\frac{E-{E}_{{\rm{g}}}}{{E}_{{\rm{U}}}}\right)$$
(1)

here α0 is a constant with units of absorption coefficient, and Eg represents the bandgap47.

DFT calculations

Density functional theory (DFT) calculations were performed using the CASTEP code48 to investigate the charge density difference, projected density of states, and local electrostatic potential at the interface between perovskites and AlOx. The exchange-correlation functional was described using the generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional49. A plane-wave basis set cutoff energy of 570 eV was employed, along with a Monkhorst-Pack k-point mesh of 1 × 2 × 2. The computational models consisted of unit cells with a 3 × 3 lateral periodicity, incorporating three octahedral layers of FAPbI3 with an exposed (100) surface, whicfh was either FAI-terminated or PbI-terminated. Slab replicas were separated by approximately 15 Å of vacuum. For geometry optimization, the Broyden–Fletcher–Goldfarb–Shannon (BFGS) algorithm was utilized. The self-consistent field (SCF) convergence criterion was set to 5 × 10−6 eV per atom, and the force tolerance was constrained to 1 × 10−2 eV Å−1.

Reporting summary

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