Introduction

Organic-inorganic metal halide perovskite solar cells (PSCs) have emerged as a leading contender in next-generation photovoltaics, owing to their remarkable progress in power conversion efficiency (PCE) over the past decade1,2,3. However, their operational stability still lags behind that of commercial crystalline silicon solar cells, posing a critical barrier to large-scale deployment4,5,6. Considerable efforts have been dedicated to addressing the intrinsic and extrinsic instability of PSCs, with strategies including additive engineering7,8, interfacial modification9,10, and low-dimensional materials11. However, nearly all of these efforts have focused on the perovskite absorber or adjacent charge transport layers, while largely neglecting the foundation of PSCs—transparent conductive oxides (TCO).

This long-standing neglect primarily arises from the widespread assumption that TCOs, such as indium tin oxide (ITO) and fluorine-doped tin oxide (FTO), are chemically inert and thus negligibly deteriorate PSCs. However, Zhao et al. recently demonstrated that volatile acidic degradation products from perovskites can permeate porous electron transport layers and react with the underlying ITO, leading to substrate corrosion12. Fortunately, outward ion diffusion/migration can be mitigated by incorporating barrier layers between the perovskite absorber and charge transport layers, thereby confining the mobile ions within the perovskite light absorber9,13,14. Nevertheless, in addition to perovskite-induced degradation, the intrinsic chemical and structural stability of the TCO itself warrants dedicated investigation—a critical issue that remains largely underexplored.

Interfacial engineering strategies can, in principle, be directly applied to TCOs to enhance their chemical and physical stability. For instance, metal oxides such as tin oxide (SnO2) or aluminum oxide (Al2O3) have been used to modify TCOs14,15. However, Al2O3 is typically prepared by the atomic layer deposition (ALD) method, which involves a more complex fabrication process14. Although lithium fluoride can also suppress ion diffusion16, it is highly hygroscopic, which can lead to moisture absorption and compromise the stability of the device. Organic materials have also been investigated for interfacial modification10,17. Yet, their limited thermal and chemical stability renders them unsuitable for long-term operation under the demanding conditions of photovoltaic devices18. Notably, stabilizing the TCO substrate is intrinsically more challenging than engineering other functional layers. As the light-incident side of the device, TCO must simultaneously offer high optical transmittance and efficient electrical conductivity19,20. These dual requirements impose stringent criteria on the interfacial materials, which must possess both a wide bandgap and favorable energy-level alignment. To meet these conflicting demands, interfacial modifiers are often applied as ultrathin films21. However, reducing the film thickness to preserve transparency and conductivity frequently undermines the mechanical integrity and functional efficacy of the interlayer, thereby limiting its ability to provide robust and durable substrate stabilization.

Here, we demonstrate that FTO is not as chemically inert as widely presumed. Under operational stressors such as illumination, elevated temperature, and electrical bias, pronounced ion diffusion occurs within the FTO bulk, undermining interfacial integrity and device stability. To mitigate this, we introduce a thermochemically robust yttrium oxide (Y2O3) interlayer, deposited via scalable thermal evaporation followed by natural oxidation. This Y2O3 interlayer forms strong interfacial bonds with FTO22 and significantly enhances interfacial adhesion energy, thereby reinforcing the intrinsic stability of FTO under stresses. Y2O3 is particularly well suited for this role due to its wide bandgap (5.6–6.0 eV), high dielectric constant, low intrinsic defect density, and high Young’s modulus23,24,25,26,27. Additionally, its optical refractive index ( ~ 2.0) closely matches that of FTO, minimizing the Fresnel reflection and preserving photon transmission.

The conformal deposition of a Y2O3 interlayer on the inherently rough FTO surface not only reinforces substrate stability but also effectively suppresses interfacial non-radiative recombination by preventing direct contact between FTO and the perovskite layer. In addition, this interlayer serves as a durable barrier against detrimental ion migration, substantially enhancing long-term operational stability. Unencapsulated devices incorporating Y2O3 exhibit negligible performance degradation after 1200 h of continuous operation, demonstrating exceptional durability. Furthermore, the versatility of this strategy is validated across multiple device configurations, achieving power conversion efficiencies of 26.48% (certified at 26.12%) for n–i–p, 26.34% for p–i–n, and 28.47% for tandem architectures. These results highlight the broad applicability and industrial promise of our approach, offering a scalable and reliable pathway toward high-efficiency, long-lifetime perovskite photovoltaics.

Results

Mechanism of interfacial reinforcement by atomically bonded Y2O3

Figure 1a illustrates the device architectures used in this study. The Y2O3 interlayer is deposited on the FTO conductive substrate. Both the regular (n–i–p) and inverted (p–i–n) configurations are shown schematically in Fig. 1a, and all-perovskite tandem devices were also fabricated on the Y2O3-coated substrate. As illustrated in Fig. 1b and Supplementary Fig. 1, we employed a two-step approach to fabricate the Y2O3 interlayer ( ~ 2 nm), combining thermal evaporation with subsequent ambient oxidation. A thin film of metallic yttrium was first deposited onto the rough FTO substrate via thermal evaporation, followed by spontaneous oxidation in ambient air to form the Y2O3 interlayer (Fig. 1c). This method offers several key advantages: (1) thermal evaporation is fully compatible with industrial-scale manufacturing, facilitating the potential commercialization of Y2O3-based interfacial engineering; (2) by evaporating metallic Y instead of directly depositing Y2O3–which has a much higher melting point–this approach significantly simplifies the fabrication process and reduce energy consumption; and (3) the initial metallic Y can react with FTO surface. As a result, a thin Y2O3 layer firmly bonded to the FTO surface is ultimately obtained (Supplementary Fig. 1d).

Fig. 1: Synthesis process of Y2O3 interlayer and FTO/Y2O3 interface reaction.
figure 1

a Schematic diagram of the PSCs structure. b Schematic illustration of the thermal-evaporation setup used to prepare a metallic Y film on an FTO substrate. Simulated atomic structures of the (c) FTO/Y2O3 and (d) FTO/SnO2 interfaces. Crystal orbital Hamilton population (COHP) analyses for (e) Sn-O bonding at the FTO/SnO2 interface and (f) Y-O bonding at the FTO/Y2O3 interface; blue lines represent COHP values and pink lines denote the integrated COHP (ICOHP).

Full size image

To gain insights into the interfacial bonding characteristics between Y2O3 and FTO, we first conducted first-principles calculations, and the computational details are presented in the Supplementary Notes. Since FTO is SnO2 doped with trace fluorine, we began by comparing the bonding strength of Y–O versus Sn–O. The calculated Y–O bond length at the FTO/Y2O3 interface is 2.008 Å (Fig. 1c), notably shorter than the 2.372 Å observed for the Sn–O bonds at the FTO/SnO2 interface (Fig. 1d). This shorter bond length implies a stronger Y–O interaction and higher intrinsic chemical stability for Y2O3. To further quantify bonding strength, we carried out crystal orbital Hamilton population (COHP) analysis. The integrated COHP (ICOHP) values reveal that the Y–O bonds exhibit an -ICOHP of 2.91 eV (Fig. 1f), exceeding the 2.50 eV of the Sn–O bonds (Fig. 1e). This higher -ICOHP value confirms that Y–O bonding is significantly stronger than Sn–O, leading to robust interfacial bonding.

To evaluate the effect of Y2O3 on interfacial adhesion, we performed first-principles calculations of the adhesion work (Wad), which quantifies the energy required to separate an interface into two distinct surfaces and serves as a metric of interfacial bonding strength28. Details of the calculation are given in Supplementary Equation 1. We calculated Wad for the native FTO/SnO2 interface, as well as for the FTO/Y2O3 and Y2O3/SnO2 interfaces. As summarized in Table 1, the Wad of the FTO/SnO2 interface is 5.88 J m−2, while that of the FTO/Y2O3 and Y2O3/SnO2 interfaces is significantly higher, at 6.86 J m−2 and 6.75 J m−2, respectively. The corresponding atomic configurations are shown in Fig. 1c, d, Supplementary Fig. 2. The increased Wad at both interfaces involving Y2O3 confirms that the incorporation of Y2O3 significantly strengthens interfacial bonding relative to the direct FTO/SnO2 contact. These findings highlight the role of Y2O3 as an effective mechanical and chemical bridge that reinforces interfacial integrity between FTO and charge-transporting layers (CTL).

Table 1 Adhesion work at various interfaces
Full size table

FTO/Y2O3 interface structure

To directly observe the interfacial reconstruction behavior induced by Y2O3, we performed high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) measurements. Focused ion beam (FIB) was used to prepare the STEM sample with structures of FTO/Y2O3/SnO2/perovskite. To investigate the bonding between Y2O3 and the FTO substrate, we deposited a 60 nm-thick Y2O3 film. Notably, to ensure complete oxidation of the metallic yttrium, the deposition process was segmented: approximately every 5 nm of Y metal deposition was followed by an ambient oxidation step. This layer-by-layer approach facilitates thorough oxidation and promotes intimate interfacial contact. To further investigate the element distribution, we performed STEM-EDS experiments on cross sections of the device. STEM-EDS mappings revealed the uniform and compact presence of the Y2O3 layer atop FTO (Fig. 2a). The distributions of tin (Fig. 2b), lead (Fig. 2c), and yttrium (Fig. 2d) were consistent with the distributions of the FTO, perovskite, and Y2O3 layers. Figure 2e is the element distribution of oxygen. The cross-sectional HAADF-STEM images in Fig. 2f revealed a dark layer sandwiched between the brighter FTO substrate and perovskite layers. The Y2O3 interlayer forms a conformal and continuous film that intimately adheres to the FTO surface. As shown in Fig. 2c, the Pb element exhibits an uneven distribution, which may be attributed to the aggregation of PbI2 at the grain boundaries7.

Fig. 2: Interface structure.
figure 2

a STEM-EDS analysis of the FTO/Y2O3/SnO2/perovskite stack. Element mappings for (b) tin, (c) lead, (d) yttrium, and (e) oxygen. f Cross-sectional high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of the FTO/Y2O3/SnO2/perovskite stack. g XRD patterns of the pristine FTO and 800 nm thick Y2O3@FTO samples. High-resolution transmission electron microscopy (HRTEM) image of the FTO/Y2O3 interface at (h) FTO region and (i) Y2O3 region.

Full size image

To explore the effect of Y on the FTO crystal planes at the interface, we first performed time-of-flight secondary ion mass spectrometry (ToF-SIMS) measurements to examine the stability of Y, as shown in Supplementary Fig. 3. After continuous heating at 85 °C for 300 h, the variation in Y content was negligible, indicating that Y2O3 possesses excellent stability and minimal elemental interdiffusion. To further investigate the influence of Y on the FTO lattice structure, we conducted XRD measurements. XRD characterization of the 2 nm-thick Y2O3 film revealed no discernible diffraction peaks attributable to Y2O3 (Supplementary Fig. 4), indicating that the film thickness is below the detection limit of the technique. To further probe potential lattice perturbations, the vicinity of the FTO (110) facet was magnified and analyzed in detail (Supplementary Fig. 5). Negligible shift in peak position was observed, implying that the Y2O3 layer predominantly resides at the outermost surface of the FTO and exhibits minimal penetration into the underlying lattice. However, when the Y2O3 film thickness was increased to 800 nm, distinct diffraction peaks became detectable in the XRD pattern (Fig. 2g). Although the 800 nm-thick Y2O3 film exhibits pronounced diffraction peaks, such as the (222) and (400) facets, we infer that the 2 nm Y2O3 layer predominantly exists in an amorphous state, as films of such minimal thickness are unlikely to develop long-range ordered crystalline structures. We performed high-resolution TEM analysis of the interfacial region FTO/Y2O3 to investigate the lattice structure across the interface. On the FTO side, a lattice spacing of 0.235 nm was identified (Fig. 2h), consistent with the (200) planes of SnO2 in FTO. As illustrated in Fig. 2i, on the Y2O3 side, we observed a lattice spacing of 0.306 nm, corresponding to the (222) planes of Y2O3.

Optoelectronic properties of Y2O3

The characterization of the optoelectronic properties of Y2O3 in this section is based on the optimized device configuration, in which the Y2O3 interlayer has a thickness of 2 nm (as discussed in the device performance section). We first characterized the optoelectronic properties of the Y2O3 interlayer using X-ray photoelectron spectroscopy (XPS). As shown in Y 3 d (Fig. 3b) and O 1 s (Fig. 3a), the surface of the deposited Y2O3 exhibits negligible oxygen vacancies, confirming the complete oxidation of metallic yttrium29,30. In contrast, a significant density of oxygen vacancies is detected on the FTO substrate surface31, as illustrated in Supplementary Fig. 6. Kelvin probe force microscopy (KPFM) was conducted to test the surface contact potential difference (CPD) in Fig. 3c (FTO) and Fig. 3d (FTO/Y2O3), which showed a slight change (dropped from 392 mV to 374 mV) in the surface CPD after Y2O3 deposition, indicating that Y2O3 plays a regulatory role in tuning the interfacial electronic properties.

Fig. 3: Properties of Y2O3 interlayer.
figure 3

XPS spectra of Y2O3, (a) O 1 s and (b) Y 3 d peaks. KPFM for (c) bare FTO and (d) Y2O3-coated FTO. e UPS of Y2O3. f REELS of Y2O3. g Band structure diagram of Y2O3, where the EVBM, ECBM, and EVAC denote the energy level of VBM, CBM, and vacuum, respectively. h Energy diagrams of the Y2O3 interlayer, SnO2 and perovskite absorber. The red arrow illustrates carrier transport to the FTO via the Y2O3 interlayer. The black arrow illustrates back-recombination of the extracted carriers and hole carriers blocked by the Y2O3 interlayer.

Full size image

The surface morphology of the prepared Y2O3 layer was examined using scanning electron microscopy (SEM) and atomic force microscopy (AFM). Supplementary Fig. 7a, b display SEM images of the FTO and Y2O3-coated FTO surfaces, respectively, revealing no discernible morphological differences between the two. To further characterize surface roughness, AFM was employed, as shown in Supplementary Fig. 7c (FTO) and 7 d (FTO/Y2O3). The root-mean-square deviation (Rq) changed negligible after the Y2O3 deposition, indicating that the Y2O3 film is conformal and does not alter the roughness of the FTO substrates. High-magnification AFM images (Supplementary Fig. 8a for FTO and 8b for FTO/Y2O3) reveal that both surfaces exhibit a similarly rough texture. To better visualize the surface roughness, two-dimensional AFM images were converted into three-dimensional ones (Supplementary Fig. 9a, b). The three-dimensional images provide a more intuitive visualization of the substantial surface roughness of the FTO substrate, highlighting the necessity of conformal deposition to prevent direct contact between the FTO and the perovskite absorber layer.

To characterize the band structure of Y2O3, ultraviolet photoelectron spectroscopy (UPS) was employed to measure the work function (WF) and valence band maximum (VBM). As shown in Fig. 3e, it indicates that the Y2O3 thin film has a WF of 3.22 eV and a VBM of 4.73 eV. Reflection electron energy loss spectroscopy (REELS) was then used to determine the band gap (Fig. 3f)32,33, which was found to be approximately 5.7 eV, consistent with values reported in the literature23,34. Such a wide bandgap typically does not compromise optical transmittance. To verify this, we compared the transmission spectra of bare FTO and FTO coated with Y2O3, as shown in the Supplementary Fig. 10. Figure 3g provides a schematic of the Y2O3 band structure with a deep VBM and shallow CBM (conduction band minimum), where the EVBM, ECBM, and EVAC denote the energy level of VBM, CBM, and vacuum, respectively. Supplementary Fig. 11 illustrates the device structure with Y2O3 serving as an interfacial layer between FTO and the SnO2 ETL (electron transport layer) in n-i-p.

In the n-i-p device, the SnO2 ETL deposited on rough FTO is responsible for conveying electron carriers generated in the perovskite absorber while simultaneously blocking holes. This necessitates that the SnO2 layer forms a uniform, dense coating over the underlying FTO substrate to prevent direct contact between the perovskite and the FTO electrode. However, the spun-cast SnO2 layers readily fail to achieve conformal coverage, leading to significant leakage19. Fabricated via thermal evaporation, the Y2O3 interfacial layer achieves an almost conformal, dense coverage on the FTO surface, effectively compensating for the leakage sites not covered by the SnO2 layer and reducing non-radiative recombination losses at the interface. Moreover, owing to its relatively shallow CBM, Y2O3 suppresses the backflow of charge carriers from the FTO electrode into the SnO2 layer or perovskite, meanwhile blocking hole carrier to FTO due to its deep VBM, thereby mitigating further carrier non-radiative recombination as shown in Fig. 3h. Although the CBM of the Y2O3 interlayer is relatively shallow, its optimized thickness of approximately 2 nm permits charge carriers to traverse the Y2O3 layer via quantum tunneling. Consequently, the presence of Y2O3 does not negatively impact charge extraction or transport21.

Figure 3h depicts energy levels of multilayered FTO, Y2O3, SnO2, and perovskite, in which energy levels of SnO2 and perovskite are based on results we have ever reported15. We anticipated that the Y2O3 nanolayer would effectively suppress the recombination of extracted carriers within the interfacial region, while minimally hindering the transport of photogenerated electrons35. To investigate the impact of Y2O3 on charge transport, we fabricated devices with structures of FTO/SnO2/Au and FTO/Y2O3/SnO2/Au to measure their J-V characteristics. As shown in Supplementary Fig. 12, the deposition of a 2 nm Y2O3 interlayer at the FTO/SnO2 interface resulted in a slight reduction in conductivity, indicating a negligible effect of the Y2O3 interlayer on the electrical performance.

To assess the morphological and electrical characteristics of SnO2 ETL grown on FTO and FTO/Y2O3 substrates, we conducted SEM and AFM analyses. According to the SEM morphology shown in Supplementary Fig. 13a, b, the surface morphology of the two samples does not differ significantly, and the subtle variation cannot be easily distinguished. Therefore, we further conducted AFM and KPFM measurements to investigate the differences in surface roughness and potential. Supplementary Fig. 13c, d demonstrate that the surface roughness of FTO/Y2O3/SnO2 is slightly lower than that of FTO/SnO2. The surface potential of FTO/Y2O3/SnO2 differs negligibly from FTO/SnO2, as shown in Supplementary Fig. 14a, b. This suggests that the Y2O3 layer beneath SnO2 does not significantly affect the electrical properties of the SnO2 ETL.

Role of the Y2O3 interlayer in inhibiting interfacial ion diffusion

The surface roughness of the FTO substrate, with an Rq of 41.3 nm (Supplementary Fig. 7c), poses a challenge for producing uniform and dense films via spin-coating, often leading to film imperfections. Effective interface materials are crucial for minimizing element migration/diffusion, which directly affects the performance and stability of PSCs36,37. Y2O3, deposited by thermal evaporation followed by natural oxidation, forms a conformal and dense layer that serves as a barrier to element migration/diffusion, playing a crucial role in enhancing the lifetime of PSCs. To assess this, we compared two samples: FTO/SnO2/perovskite and FTO/Y2O3/SnO2 /perovskite. Both samples were subjected to heating at 85 °C for 300 h and analyzed using ToF-SIMS. The results show that fluorine and oxygen migration/diffusion into the SnO2 ETL and perovskite layer were significantly suppressed in the FTO/Y2O3/SnO2/perovskite samples (Figs. 4d, e) compared to the FTO/SnO2/perovskite ones (Figs. 4a, b). This demonstrates that Y2O3 effectively suppresses fluorine and oxygen element segregation, preserving the integrity of the FTO electrode and hence improving the overall performance and stability of PSCs.

Fig. 4: Role of the Y2O3 interlayer in inhibiting interfacial ion diffusion.
figure 4

a–c Variation of F, O, and I content within the FTO/SnO2/perovskite configuration after 300 h of heating at 85 °C. d–f Variation of F, O, and I ion content within the FTO/Y2O3/SnO2/perovskite configuration under the same conditions. g Ion diffusion in the FTO/SnO2/perovskite configuration. h Ion diffusion in the FTO/Y2O3/SnO2/perovskite configuration.

Full size image

To further illustrate the impact of Y2O3 on ion diffusion in FTO, KPFM was employed to assess the evolution of surface potential in FTO/SnO2 and FTO/Y2O3/SnO2 samples aging at 85 °C for 300 h. In the FTO/SnO2 sample, the surface potential dropped from 464 mV (Supplementary Fig. 14a) to 368 mV (Supplementary Fig. 15a), a decrease of 96 mV. In contrast, the surface potential of the FTO/Y2O3/SnO2 sample only decreased from 463 mV (Supplementary Fig. 14b) to 427 mV (Supplementary Fig. 15b), a much smaller drop of 36 mV. This suggests that fluorine and oxygen segregation significantly affect the energy level of the SnO2 ETL, affecting electron transfer and device performance. UPS (Supplementary Fig. 16) and optical transmittance (Supplementary Fig. 17) were used to further characterize this variation. After continuous heating at 85 °C for 300 h, both the VBM and optical transmittance of all samples exhibited negligible changes, indicating that the CBM also remained essentially unchanged. In contrast, the Fermi level (EF)of the control sample decreased markedly, whereas that of the Y2O3-modified sample showed almost negligence. This result is consistent with the KPFM measurements, further confirming that the introduction of the Y2O3 layer effectively stabilizes the interfacial electronic structure.

Additionally, significant iodine from the perovskite layer migration/diffusion into the SnO2 ETL and FTO substrate was observed in the control samples (Fig. 4c), whereas minimal iodine was detected in the FTO/Y2O3/SnO2 ones (Fig. 4f). In addition to suppressing the migration of oxygen, iodine, and fluorine through the interface, Y2O3 has also been tested and verified for its effect on other elements. We also examined the diffusion of Sn and Pb elements (Supplementary Fig. 18), and both exhibited negligible changes after aging.

We illustrated the role of Y2O3 using a schematic diagram, as shown in Figs. 4g, h. In the PSCs without Y2O3 interlayer, ion diffusion/migration occurs at the interface between FTO and ETL. Element such as iodine in the perovskite can migrate/diffuse across the interface, corroding the FTO electrode, while oxygen and fluorine from the FTO electrode can also migrate/diffuse through the interface and affect the perovskite light-absorbing layer, as shown in Fig. 4g. This process is effectively mitigated after introducing the Y2O3 interface layer, as shown in Fig. 4h. Furthermore, the spin-coated SnO2 layer fails to form a dense coating on the FTO substrate, direct contact between FTO and the perovskite absorber layer occurs, leading to the formation of numerous leakage sites, as shown in Fig. 4g. However, with the bonding of Y2O3, these sites are effectively covered, preventing direct contact between FTO and the perovskite layer and thereby reducing leakage current. Overall, the inclusion of Y2O3 offers multiple benefits: it reduces non-radiative recombination at the interface, significantly mitigates element segregation, and prevents leakage current formation, all of which contribute to enhanced device efficiency and longevity.

Photovoltaic performance

By incorporating an optimized Y2O3 interlayer in n-i-p PSCs, we achieved a power conversion efficiency (PCE) of 26.48% (with a certified PCE of 26.12%, Supplementary Fig. 19), notably higher than the control with a PCE of 24.38%. To verify whether our approach is universal or not, we also fabricated a p-i-n PSC with Y2O3. The champion device yielded a PCE of 26.34%, with a fill factor (FF) of 86%, an open-circuit voltage (VOC) of 1.151 V, and a short-circuit current density (JSC) 26.61 mA cm−2. The specific performance parameters are given in Supplementary Fig. 20b, with the schematic structure shown in Supplementary Fig. 20a. As illustrated in Fig. 5a, the most significant improvements with Y2O3 in the n-i-p structured PSCs originated from the enhancement of the VOC and FF, while the JSC showed a modest increase. Figure 5b demonstrates that the stabilized efficiency at the maximum power point for the Y2O3-enhanced cell is approximately 26.37%, substantially higher than the control cell with 23.78%. To ensure the reliability of JSC, we measured the external quantum efficiency (EQE), which showed integrated current densities of 25.72 mA cm−2 for the Y2O3-enhanced cells and 25.58 mA cm−2 for the control ones, respectively (Fig. 5c), consistent with JSC values from J-V curves. Statistics of performance metrics indicate that a 2 nm thick Y2O3 layer gives rise to the optimal enhancement (Supplementary Fig. 21).

Fig. 5: Characterization of device performance.
figure 5

a J-V curves of the n-i-p devices. b Stabilized power output (SPO) of the n-i-p devices at the maximum power point (MPP). c EQE spectra of the n-i-p devices. d J-V curves of the tandem devices. e SPO of the tandem devices at the MPP. f EQE spectra of the bottom subcell and top subcell for the tandem devices. g Temporal evolution of PCEs of the unencapsulated devices under MPP tracking at 35 °C in a N2 atmosphere.

Full size image

To further validate the versatility of the Y2O3 interlayer, we integrated it into all-perovskite tandem solar cells. Remarkably, the resulting tandem devices achieved a power conversion efficiency (PCE) exceeding 28%, as shown in Fig. 5d. The stabilized power output also surpassed 28%, as presented in the steady-state efficiency measurement in Fig. 5e. Figure 5f displays the EQE spectra of both the wide-bandgap and narrow-bandgap subcells, revealing closely matched integrated JSC. The successful application of Y2O3 in tandem devices further highlights its broad compatibility and underscores its strong potential for commercialization.

To elucidate the origins of the observed performance enhancement, we first measured the dark current density of the devices. As shown in Supplementary Fig. 22, the Y2O3-modified cells exhibit a substantially lower dark current, indicative of suppressed leakage pathways38. We then performed Mott-Schottky analysis (Supplementary Fig. 23), which reveals an increase in built-in potential (Vbi) from 0.83 V in the control to 0.95 V in the Y2O3-treated devices, a change that directly contributes to higher VOC39. To further substantiate the reduction in non-radiative recombination, we measured the quasi-Fermi level splitting (QFLS) of different samples40,41, as shown in Supplementary Fig. 24. Samples incorporating the Y2O3 interlayer exhibited a noticeably greater QFLS, indicating reduced non-radiative recombination and enhanced optoelectronic quality. Confocal photoluminescence (PL) mapping measurements in Supplementary Fig. 25 demonstrating the same conclusions42. This observation was further corroborated by KPFM, which revealed consistent trends in surface potential distribution. KPFM (Supplementary Fig. 26) shows a more uniform and reduced surface potential for perovskite films with the Y2O3 interlayer, reflecting diminished non-radiative recombination and underpinning improvements in both VOC and FF43.

To further investigate the effect of Y2O3 on the suppression of interfacial non-radiative recombination, we first measured the time-resolved photoluminescence (TRPL) and steady-state photoluminescence (PL) spectra of FTO/perovskite and FTO/Y2O3/perovskite samples. As shown in Supplementary Fig. 27, the Y2O3-modified sample exhibits higher PL intensity and a longer carrier lifetime, indicating that interfacial recombination is effectively suppressed. This improvement can be attributed to the introduction of Y2O3, which prevents direct contact between the perovskite and the FTO substrate due to the incomplete coverage of the SnO2 electron transport layer, thereby mitigating interfacial non-radiative recombination35. In addition, we characterized the recombination resistance via electrochemical impedance spectroscopy (EIS). As shown in Supplementary Fig. 28, the Y2O3-modified sample exhibits a higher recombination resistance, further confirming that the incorporation of Y2O3 increases the resistance to carrier recombination, thereby reducing non-radiative losses44,45. These results consistently demonstrate that Y2O3 plays a crucial role in passivating interfacial defects and improving the optoelectronic quality of the device.

In the inverted structure, the adsorption of (4-(7H-dibenzo[c,g]carbazol-7-yl)butyl)phosphonic acid (4PADCB) on the substrate depends on the density of –OH groups46. Since the surface of Y2O3 contains fewer –OH groups, we investigated its influence on small-molecule adsorption by measuring the change in surface potential of FTO/4PADCB and FTO/Y2O3/4PADCB using KPFM47. As shown in Supplementary Fig. 29, the surface potential difference between the two samples is slight. Moreover, prior to the deposition of small molecules, the surface of Y2O3 undergoes ozone treatment to increase the –OH group density46. Therefore, the Y2O3 interlayer has minimal impact on the adsorption of small molecules.

To assess the impact of Y2O3 on the thermal stability of PSCs, we conducted a continuous 430-h test at 85 °C. As shown in Supplementary Fig. 30, the target devices retained over 80% of their initial PCEs after aging, whereas the control dropped to around 70%, indicating that Y2O3 significantly improves the thermal stability of PSCs. To further evaluate the operational stability, we recorded the temporal evolution of their efficiencies at the maximum power point (MPP). As shown in Fig. 5g, after over 1200 h, the PCEs of the control devices decreased to approximately 80% of their initial ones with considerable fluctuations during operation. In contrast, the target group maintained nearly 98% of their initial PCEs under the identical conditions. This indicates that Y2O3 effectively enhances the operational stability of the devices. To eliminate the impact of spiro-OMeTAD instability, we replaced it with poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA) as the hole transport layer (HTL) for the stability measurement48.

Further analysis involved X-ray diffraction (XRD) of perovskite films aged at 85 °C for 300 h (Supplementary Fig. 31). The perovskites maintained their crystal orientation with minimal phase changes, suggesting that ion diffusion does not severely affect the perovskite phase over short-term aging. Confocal PL mapping (Supplementary Fig. 32) revealed that the fluorescence intensity of the target sample remained strong, while the control significantly diminished. This indicates that Y2O3 helps mitigate the negative effects of element segregation on the optoelectronic performance of perovskite films.

Discussion

Our findings reveal that under operating conditions, FTO undergoes a degradation issue, which critically undermines device stability. To address this hidden vulnerability, we introduce a universal and scalable interface engineering strategy that forms a conformal Y2O3 interlayer on the rough FTO via thermal evaporation followed by room-temperature oxidation. This interlayer not only establishes strong interfacial bonding with FTO—marked by enhanced adhesion energy—but also effectively suppresses ion diffusion, thereby stabilizing the substrate. As a result, we achieve unprecedented operational durability, with unencapsulated devices exhibiting negligible performance degradation after 1200 h of continuous operation. Furthermore, the Y2O3-modified FTO proves broadly applicable, enabling high efficiencies across diverse device architectures—including 26.48% (certified 26.12%) for n–i–p, 26.34% for p–i–n, and 28.47% for all perovskite tandem devices. These results highlight the overlooked importance of substrate engineering in PSCs and position our strategy as a robust and commercially relevant pathway toward durable, high-efficiency perovskite photovoltaics.

Methods

Materials

The SnO2 (15 wt.% colloidal dispersion tin (IV) oxide) was purchased from Alfa Aesar. Hydrogen peroxide 30% aqueous solution (H2O2) was purchased from Sinopharm Chemical Reagent Co., Ltd. Methylamine hydrochloride (MACl, ≥99.5%), Poly[bis(4-phenyl)(2,4,6- trimethylphenyl)amine] (PTAA), C60, BCP and oleylammonium chloride (OACl) were purchased from Xi’an Yuri Solar Co., Ltd. Lead (II) iodide (PbI2, 99.99%), 4-Isopropyl-4’-methyldiphenyliodonium Tetrakis (pentafluorophenyl) borate (TPFB), (4-(7H-dibenzo[c,g]carbazol-7-yl)butyl)phosphonic acid (4PADCB) and potassium iodide (KI) was purchased from TCI Shanghai (China). N,Ndimethylformamide (DMF, 99.8%), dimethyl sulfoxide (DMSO, 99.9%), isopropanol (IPA, 99.8%), acetonitrile (CAN, 99.8%), chlorobenzene (CB, 99.8%), bis(trifluoromethane) sulfon imide lithium salt (Li-TFSI), and 4-tert-butylpyridine (tBP) were purchased from Sigma-Aldrich. Formamidinium iodide (FAI, ≥99.5%), 2,2′,7,7′-tetrakis-(N, N-di-p-methoxyphenylamine)−9,9′spirobifluorene (Spiro-OMeTAD, ≥99.5%), SnI2, CsI, PbBr2 were purchased from Advanced Election Technology CO., Ltd. Yttrium powder (99.9%, metals basis) was purchased from Aladdin Scientific Corp. All chemicals were used as received without any other purification.

Fabrication

Preparation of Y2O3 thin films

The FTO was sequentially cleaned using glass cleaner, deionized water, acetone, and isopropanol with ultrasonic treatment for 15 min each. Subsequently, a layer of metallic yttrium is evaporated on FTO substrates at a controlled rate of 0.01 Å/s in a high-vacuum environment with a pressure of approximately 3 × 10−4 Pa. Then the freshly deposited metallic yttrium film is exposed to air at room temperature for about 10 min. This exposure allows the yttrium to naturally oxidize, forming Y2O3.

Device fabrication

The commercial SnO2 colloidal precursor stock solution was diluted with deionized water and H2O2 at a volumetric ratio of 1:4:1 (stock solution: deionized water: H2O2 = 1:4:1). The Y2O3-coated FTO was treated with ultraviolet ozone for 15 min. Then the SnO2 precursor was spin-coated onto the FTO or Y2O3-FTO at 4000 rpm for 30 s. The spin-coated substrate was annealed on a hot plate at 150 °C for 30 min in air. After cooling to room temperature, the FTO was transferred to glovebox to fabricate perovskite. PbI2 precursor solution (1.5 M PbI2 and 0.04 M KI in a mixed solvent of DMF and DMSO with a volume ratio of 900:100) was spin-coated onto SnO2 at 1500 rpm for 30 s, then annealed at 70 °C for 1 min. For perovskite film deposition, a solution of FAI: MACl (90 mg: 15.8 mg in 1 ml IPA) was spin-coated onto the PbI2 film at 2000 rpm for 30 s. The substrate was then transferred on a hot plate and heated at 150 °C for 13 min with 35% RH. OACl (1 mg in 1 mL IPA) was deposited on perovskite films at 4000 rpm for 30 s as a passivation layer. After the passivation layer deposition, the hole-transport layer was subsequently deposited on top of the perovskite/OACl by spin coating at 3000 rpm for 20 s using a spiro-OMeTAD/CB solution, which consisted of spiro-OMeTAD (72.3 mg), tBP (28.8 μL), Li-TFSI/ACN (17.5 μL, 520 mg mL−1), and CB (1 mL). Finally, 60 nm Au was deposited by thermal evaporation. For the stability test of the device, PTAA doped with TPFB was employed as a replacement for spiro-OMeTAD as the HTL. The concentration of PTAA was 30 mg mL−1, with a weight ratio of PTAA to TPFB of 10:1.

For p-i-n solar cells, the Y2O3 (0.5 nm)-coated FTO was treated with ultraviolet ozone for 20 min. The samples were then transferred to an N2-filled glove box. Next, 4PADCB solution (0.3 mg mL−1 in ethanol) was spin-coated onto FTO at 3000 rpm for 30 s following by annealing at 100 °C for 10 min. After cooling to room temperature, PbI2 precursor solution (1.5 M in a mixed solvent of DMF and DMSO with a volume ratio of 900:100) was spin-coated onto 4PADCB at 1500 rpm for 30 s, then annealed at 70 °C for 1 min. For perovskite film deposition, a solution of FAI: MACl (90 mg: 15.8 mg in 1 mL IPA) was spin-coated onto the PbI2 film at 2000 rpm for 30 s. The substrate was then transferred to a hot plate and heated at 140 °C for 10 min with 35% RH. Then the PDAI2 solution in isopropanol was deposited on the perovskite at 3000 rpm for 30 s following by annealing at 100 °C for 5 min. Finally, 20 nm of C60, 7 nm of BCP, and 100 nm of Cu were sequentially deposited by thermal evaporation.

Tandem devices fabrication

Narrow-bandgap (NBG) FA0.7MA0.3Pb0.5Sn0.5I3

A 2.4 M precursor solution was prepared by dissolving PbI2, SnI2, MAI, and FAI in a mixed solvent of DMF and DMSO (v/v = 3:1). To suppress Sn2+ oxidation and create a Sn-rich environment, SnF2 (10 mol% relative to SnI2) was added. The mixture was stirred at room temperature for 1 h, followed by filtration through a 0.22 μm PTFE membrane before use.

Wide-bandgap (WBG) FA0.8Cs0.2Pb(I0.6Br0.4)3

The 1.2 M WBG precursor solution was prepared by dissolving CsI, FAI, PbBr2, and PbI2 in DMF and DMSO (v/v = 4:1). The solution was stirred at 60 °C for 1 h and filtered through a 0.22 μm PVDF membrane before deposition.

Fabrication of all-perovskite tandem solar cells

The Y2O3 (0.5 nm)-coated FTO was treated with ultraviolet ozone for 20 min. NiOx nanocrystal layers (10 mg mL−1 in water) were spin-coated at 3000 rpm for 30 s and annealed at 100 °C for 10 min. Then, a Me-4PACz self-assembled monolayer (0.3 mg mL−1 in ethanol) was deposited by spin-coating at 3000 rpm for 30 s and annealed at 100 °C in a nitrogen atmosphere. WBG perovskite films were fabricated by spin-coating 50 μL of the precursor solution at 5000 rpm for 60 s. At 40 s into the process, 300 μL of diethyl ether was dropped. Substrates were then annealed at 50 °C for 2 min and 100 °C for 10 min. After cooling, a post-treatment with PDAI2 (2 mg mL−1 in IPA) was performed by spin-coating at 4000 rpm for 30 s and annealing at 100 °C for 5 min. A C60 layer (18 nm) was thermally evaporated onto the WBG film, followed by the deposition of a 20 nm SnOx layer via atomic layer deposition (ALD) and 0.8 nm of Au via thermal evaporation. The NBG perovskite layers were deposited via a two-step spin-coating process: (1) 1000 rpm for 10 s (acceleration: 200 rpm s−1), and (2) 4000 rpm for 40 s (acceleration: 1000 rpm s−1). During the second step, 400 μL of chlorobenzene was dripped 30 s after spin initiation. Films were annealed at 100 °C for 10 min. Post-treatment was performed by spin-coating EDAI2 (0.5 mg mL−1 in IPA) at 4000 rpm for 30 s, followed by annealing at 100 °C for 5 min. Finally, C60 (20 nm), BCP (7 nm), and Cu (100 nm) were deposited sequentially via thermal evaporation to complete the tandem device.

Characterization of perovskite films and devices

Device J-V performance characterization

Measured under AM 1.5 G solar simulation (AAA grade, 450 W xenon lamp, Newport 94043 A) using a Keithley 2400 source meter. The light intensity was calibrated with a Newport reference solar cell (91150-KG5). J-V measurements were conducted in a nitrogen atmosphere with a step of 20 mV, ranging from −0.2 to 1.2 V, followed by a reverse scan from 1.2 to −0.2 V, with a duration of 20 ms.

Film characterization

Cross-sectional and top-view SEM images were taken using a JSM 6700 F field emission scanning electron microscope. AFM and KPFM images of the perovskite films were obtained using an atomic force microscope (Dimension ICON) in peak force mode. UV-Vis absorption spectra were measured with a Shimadzu UV-1700 spectrophotometer. X-ray diffraction (XRD) and grazing incidence X-ray diffraction (GIXRD) spectra were obtained using a HyPix-3000 diffractometer. Photoluminescence (PL) and time-resolved photoluminescence (TRPL) spectra were recorded using a Delta Flex spectrofluorometer (HORIBA). X-ray photoelectron spectroscopy (XPS) was performed with a Thermo Scientific (Escalab 250Xi) instrument, with binding energy shifts calibrated using the C1s peak at 284.8 eV. The EQEEL measurement uses a Keithley 2901 source meter to collect current density and drive voltage, and an integrating sphere with the LQ-100 system (Enlitech Co., Ltd.) to gather the luminescence signal. The IPCE spectrum is measured using the QE/IPCE system from Enli Technology Co., Ltd. TOF-SIMS (Tescan AMBER) spectrometer was used for depth profile analysis of perovskites with 30 KV and 100 pA. Transient absorption spectra were measured using an ultrafast transient absorption spectrometer (HARPIA, Light Conversion). TEM measurements were characterized with JEM-ARM300F2(GRAND ARM2). The characterization of quasi-Fermi level splitting (QFLS) was conducted by QFLS-Maper (Enli Technology Co., Ltd).