Introduction

Multi-junction tandem solar cells (TSCs) have garnered research attention for their competitive advantages in overcoming the Shockley–Queisser efficiency limitations of single-junction solar cells.^1-3 Perovskite solar cells (PSCs) have tunable band gaps, long charge carrier diffusion lengths, superior light collection efficiency, and facile fabrication processes, contributing to their impressive efficiency rate of 26.1 %.^{4, 5} These qualities make PSCs ideal candidates for top cells in TSCs.⁶ The certified efficiency of perovskite-on-silicon 2-terminal TSCs stands at 33.9 %.⁷ However, the issue of light-induced phase segregation (LIPS) in wide-band gap (>1.65 electron volts, eV) iodide/bromide (I/Br) mixed perovskite absorbers poses a challenge to the power conversion efficiency (PCE) and stability of perovskite-based TSCs.^8-10 Substantial efforts have been dedicated to mitigating halide segregation through crystallization control, defect passivation, and strain engineering methods.^11-14 Despite some encouraging progress, the photostability of wide-band gap PSCs remains unsatisfactory.¹⁵

Moreover, the solution processing and the low formation energy of perovskites inevitably lead to numerous defects formed at both the bulk and interfaces of the perovskite layer.^16-18 These defects can act as non-radiative recombination centers, significantly impeding carrier transport.^19-21 This ultimately results in a substantial deficit in the open-circuit voltage (V_oc), particularly at the interface between perovskites and fullerene electron transport layers (ETLs) in inverted PSCs.^{22, 23} Passivating these defects is paramount in realizing the high efficiency and stability of PSCs.²⁴ Various strategies have been devised to tackle these challenges, including composition tuning, employing low-dimensional perovskites, and optimizing interfaces.^25-27 However, many of these approaches rely on the spin-coating method, which is unsuitable for treatment on textured surfaces and may not be compatible with large-area substrates.

Atomic layer deposition (ALD) has found widespread application as a highly effective method for depositing large-area thin films.²⁸ It has been utilized in fields ranging from organic light-emitting diodes (OLEDs) to silicon and thin-film solar cells.^{29, 30} When applied between the surface of the perovskite layer and the transport layer, AlO_x deposited via ALD has proven to be efficacious in enhancing the performance of PSCs.^{30, 31} While most studies focus on methylammonium (MA)-based perovskites,³² the working mechanism and functions of these AlO_x layers remain unclear, particularly for formamidinium (FA)-based perovskites.

Here, a facile and efficient strategy is employed for precisely modulating the perovskite by incorporating AlO_x deposited by ALD on the top interface. Al³⁺ ions were found to be infiltrated in the perovskite layer and interact with halide ions. The treatment contributed to realizing better-matched energy levels, suppressed ion migration and minimized interfacial carrier losses simultaneously. Additionally, the self-encapsulation effect of this dense interlayer can inhibit volatile ion overflow at high temperatures and improve light and thermal stability. Consequently, the ALD-AlO_x modification could significantly improve the PCE of wide-band gap PSCs from 19.32 % to 21.80 % and enhance their photothermal stability. More importantly, a monolithic perovskite-silicon TSC using AlO_x-modified perovskite achieved a PCE of 28.50 % with good stability. The resulting 1.55 eV PSC and module also achieved a PCE of 25.08 % (0.04 cm²) and 21.01 % (aperture area of 15.5 cm²), respectively, proving the universality of the strategy.

Results and Discussion

The ultrathin AlO_x film was fabricated by atomic layer deposition (ALD) with trimethylaluminum (TMA) and H₂O. As shown in Figure 1, TMA and H₂O are alternately exposed to a substrate in a repeated TMA-pulse-purge-H₂O-pulse-purge sequence, and we proposed a growth mechanism of ALD AlO_x on top of perovskite (Figure 1).^{32, 33} During the first half-cycle (equation 1), TMA reacts with the perovskite surface by interacting with the organic cation (MA⁺ or FA⁺), which weakens the hydrogen bonds between the organic cation and I⁻ of the perovskite, leading to a breakdown of the perovskite framework on the surface. The adduct comprising of X₃Pb-Al(CH₃)-PbX₃ as the key intermediates was generated accompanied by the release of CH₄ and FA or MA, and then, X₃Pb-Al(CH₃)-PbX₃ reacted with the H₂O molecule during the subsequent half-cycle and produced the −OH surface sites (X₃Pb-Al(OH)-PbX₃) (equation 2), which are considered to be the most ubiquitous chemisorption sites for TMA and necessary to promote the growth of AlO_x. With the increased number of ALD cycles (equations 3 and 4), a dense AlO_x film was formed on the perovskite. The final thickness of AlO_x can be precisely controlled by the number of ALD cycles.

We verified the ultrathin nature of the AlO_x interlayer with high-resolution scanning transmission electron microscopy (HR-STEM). As shown in Figure 2a and S1, the STEM images and energy-dispersive X-ray spectroscopy (EDS) mapping clearly outline the perovskite/AlO_x/C₆₀/SnO₂ top contact structure, identifying the presence of a ~18 nm C₆₀ layer and a ~10 nm SnO₂ layer. Interestingly, Al³⁺ ions are found to be diffused to the bulk and throughout the whole perovskite layer (Figure S2), which are supposed to exist in the grain boundaries rather than dope within the perovskite lattice due to the much smaller ionic radius than Pb²⁺ (53.5 pm vs. 119 pm).^{34, 35} There is no noticeable change in the X-ray diffraction (XRD) of the perovskite film before and after modifying with ALD AlO_x (Figure S4), confirming our speculation. In addition, the modified perovskite films in atomic force microscope (AFM) and scanning electron microscopy (SEM) exhibited reduced roughness and increased grain size, indicating that the surface morphology was improved by AlO_x treatment (Figure S5–S6). The infiltrated Al³⁺ ions at the grain boundaries and perovskite surface would passivate the antisite PbX₃⁻ (X means I or Br) defects and suppress the ion migration pathway, facilitating the phase stability of bromine-rich wide-band gap perovskite. Moreover, the dense interlayer of AlO_x not only acts as an inner encapsulation but also regulates the energy level arrangement at the perovskite/C₆₀ interface (Figure 2b).

The energy band structure of perovskite (PVSK) and PVSK/AlO_x was determined via ultraviolet photoelectron spectroscopy (UPS) and optical band gap (E_g) values. As shown in Figure 3a and Figure S7, the Fermi levels (E_F) of PVSK film are calculated from the equation of E_F=h$$ −E_cut-off (h$$ =21.2 eV), which upshifts from −4.27 to −4.19 eV after capping with AlO_x. The conduction band minima (CBM) of the PVSK/AlO_x is evaluated by the valence band maxima (VBM) and E_g values, which is determined to be −4.09 eV, bringing it closer to the CBM of C₆₀ than PVSK (−4.10 eV). Kelvin probe force microscopy (KPFM) was carried out to explore their surface potentials, as shown in Figure 3b–c and S7. The surface of PVSK/AlO_x exhibits an average surface potential of 787 mV, which is higher than that of bare PVSK film (557 mV), consistent with the results of the UPS measurements above. The lower offset between E_F and CBM of PVSK/AlO_x indicates its stronger n-type characteristics, which could suppress the trap-assisted recombination loss and promote charge extraction from perovskite to ETL efficiently, resulting in an improved V_oc.^{36, 37}

To unravel the mechanism behind the impact of the AlO_x interlayer, we conducted DFT calculations to explore the interaction between C₆₀ and the perovskite surface. As depicted in Figure 3d–e, the C₆₀ and PVSK/AlO_x surface exhibits higher binding energy −20.3 kcal mol⁻¹ than that of C₆₀ and PVSK (−15.2 kcal mol⁻¹). The larger binding energy suggests tighter packing between C₆₀ and PVSK/AlO_x, improving the interfacial contact between perovskite and C₆₀³⁸ and contributing to the enhanced thermal stability of the device (discussed below). Moreover, a surface dipole can increase the local work function and hinder efficient electron extraction.³⁹ The z-component of the dipole moment (D_z) is calculated to be maximal for the PVSK/C₆₀ system (4.10 Debye) and is reduced to 2.31 Debye after the insertion of AlO_x interlayer (Figure S9), consistent with the trend of work functions measured by UPS.⁴⁰

X-ray photoelectron spectroscopy (XPS) was provided to further insight into the chemical modification of the perovskite. As shown in Figure 3f, the presence of AlO_x is verified by the additional Al 2p peak at 74.8 eV for PVSK/AlO_x.⁴¹ The Br 3d spectrum of PVSK/AlO_x slightly shifted to higher binding energies, which could be a result of the electron transfer between the Al³⁺ ion and Br⁻ in perovskite.³⁵ Furthermore, both Pb 4f_7/2 and Pb 4f_5/2 peaks (Figure 3g) shift towards lower binding energy values by 0.4 eV after the AlO_x modification, which is attributed to the passivation of Pb-rich surface defects with lattice oxygen of AlO_x. Moreover, the two weak signals of Pb 4f_7/2 at ∼136.5 eV and Pb 4f_5/2 at ∼141.4 eV in the pristine PVSK correspond to the metallic Pb⁰ (Figure 3g), which disappear after the capping with AlO_x, indicating that AlO_x significantly suppresses the formation of metallic Pb⁰ on the perovskite surface.⁴² The O 1s of bare AlO_x can be de-convoluted into the Al−O bond at 532.5 eV and −OH bond at 531.8 eV⁴³ (Figure 3h), and the characteristic peaks of Al−O shift towards higher binding energy from 532.5 eV to 533.0 eV, which further confirmed the interaction between AlO_x and perovskite. The interaction was also demonstrated by the peak shifts of I 3d in the perovskite film before and after AlO_x treatment (Figure S10).

where V_TFL represents the transition voltage between ohmic and trap-filling regions, e is the elemental charge, and L is the perovskite film thickness. As determined from the density–voltage (J−V) curve of the electron-only devices in the dark (Figure 3k), the corresponding trap densities (N_t) were determined to be 4.87×10¹⁵ and 2.30×10¹⁵ cm⁻³ for the devices based on PVSK and PVSK/AlO_x, respectively. The reduced N_t of the perovskite film capping with AlO_x is ascribed to its passivation capability on the perovskite surface and grain boundaries,⁴⁸ consistent with the analysis of XPS mentioned above.

The single-junction wide-band gap perovskite solar cells were first fabricated with and without the AlO_x layer. The device configuration is ITO/MeO-2PACz/Perovskite/AlO_x/C₆₀/SnO₂/Ag (Figure 4a), and the J−V curves were measured under one-sun illumination. The optimized J−V curves and data of control and target PSCs are shown in Figure 4b and Table S3. The devices incorporating AlO_x showed the highest PCE of 21.80 % with a short-circuit current density (J_sc) of 22.59 mA cm⁻², V_oc of 1.181 V, and fill factor (FF) of 81.72 %. By contrast, the control cell presented a much lower PCE of 19.32 % due to the lower V_oc of 1.126 V and FF of 78.62 % (Figure S12). The integrated J_sc derived from external quantum efficiency (EQE) spectra (Figure 4c) are 21.26 and 22.00 mA cm⁻² for the control and target device, respectively, approximate to their measured J_sc. The steady-state power output (SPO) at the maximum power point was carried out to verify the performance of modified PSCs (Figure S13–S14). The PCE of the optimized device was further verified by the stabilized power output measurements at the maximum power point (V_max=1.00 V for target, V_max=0.96 V for control) for 200 s, which shows a stabilized PCE of 21.23 % and 18.53 % for the target and control device, respectively. Moreover, the target cells exhibit good reproducibility with an average PCE of 21.17 % (Figure 4d), much better than the control device (19.40 %). The improved performance of the target cell with the AlO_x layer mainly originated from the increased FF and V_oc (Figure S15). Accordingly, the built-in potential (V_bi) of the devices is derived from the capacitance-voltage curves, which were characterized to directly compare by the Mott–Schottky analysis.⁴⁹ The target device with an AlO_x layer exhibits a larger V_bi of 1.07 V than that of the control cell (1.05 V) (Figure 4e), indicating a stronger driving force of the target device for separating photogenerated charge carriers, which is facilitated to achieve a higher V_oc. We ascribe the superior V_bi to the better-matched energy level alignment and defect passivation at the perovskite/ETL interface. Moreover, the leakage currents from the dark J−V curves of PSCs were performed (Figure 4f). The reverse leakage current and ideal factor (m) of the target device with the AlO_x layer is smaller than that of the control, indicating the minimized shallow defect states and trap-assisted Shockley–Read–Hall recombination.⁵⁰ The results are in accordance with the measurements of electrochemical impedance spectroscopy. The Nyquist plots for different PSCs with and without the AlO_x layer were measured at 1.0 V, ranging from 10⁶ to 1 Hz in the dark. As shown in Figure 4g, the charge transport resistance of the target device (934 Ω) was much lower than that of the control cell (1692 Ω), further confirming the adequately suppressed carrier recombination at the perovskite/C₆₀ interface or in bulk.

To investigate the universality of our strategy in enhancing inverted device performance, we have also extended the application of the AlO_x treatment to the devices utilizing perovskite absorber materials with band gaps of 1.55 eV. The detailed device fabrication process of the optimized PSCs is presented in the Supporting Information. As shown in Figure 4h, the optimized device based on the AlO_x modification yields an efficiency of 25.08 % with a high V_oc of 1.173 V, an FF of 84.83 % and a J_sc of 25.20 mA cm⁻². Additionally, we successfully fabricated a PSC mini-module with an aperture area of around 15.5 cm² (The geometric fill factor is 92.4 %) and tested its J–V characteristics (Figure 4i). The PCE of the mini-module with 8 sub-cells reached 21.01 %, with a V_oc of 9.474 V, a J_sc of 2.833 mA cm⁻² and an FF of 78.27 %.

The stability of perovskite-based TSCs is largely dependent on wide-band gap perovskites. Therefore, we performed in situ photoluminescence (in situ PL) to evaluate the stability of wide-band gap perovskites under light illumination conditions. As shown in Figure 5a, b and S15, the PL peak of the control film was significantly redshifted after continuous illumination (laser) for 150 seconds. Simultaneously, a new PL peak with lower photon energy (~1.59 eV), indicative of halide segregation, gradually appeared in the PL spectrum of the control sample within 150 seconds. Furthermore, the increase in the full width at half maximum (FWHM) (Figure 5d) and the reduction in PL intensity (Figure 5c) indicated the formation of new iodine-rich and bromine-rich phases.⁵¹ In contrast, the target film showed consistent PL peak positions, PL intensity, and FWHM values, suggesting that the infiltrated Al³⁺ effectively impedes ion migration pathways and maintains the structural stability of the perovskite crystal. The long-term stability of wide-band gap PSCs was further explored. As depicted in Figure 5e, the target PSC can sustain over 97 % of its initial PCE after being stored in N₂ for 3500 hours. In comparison, the PCE of the control one dropped to 54 % of the initial value. The moisture and oxygen stability of the PSCs were conducted under a relative humidity (RH) of 35±5% (Figure S17). The target PSC maintains approximately 90 % of its original PCE after storing over 1300 hours. In contrast, the control PSC declines to 43 % of its original PCE. Moreover, we further investigated the light stability under continuous light soaking (white LED, 100 mW cm⁻²) in a glove box. The target PSCs with the AlO_x layer can maintain over 87 % of their original PCE after 1560 hours, while the PSC based on pristine PVSK dropped to 57 % of its premier value (Figure 5f). We explored the thermal aging test under one-sun irradiation to accelerate the aging test. As shown in Figure 5g, the unencapsulated device modified by the AlO_x layer could sustain 78 % of its original PCE after being stored at 65 °C in N₂ under continued light soaking for 672 hours, while the PCE of the control cell retained only 38 % of its initial value. The thermal stability of the perovskite films was also tracked using XRD measurements. As shown in Figure 5h and 5i, the diffraction peak around 12.3° is assigned to the characteristic peak of PbI₂, which was observed for the pristine perovskite film after 10 days. The diffraction intensity sharply decreases along with the storage time due to the phase segregation of wide-band gap perovskite. In contrast, no obvious PbI₂ signals existed until 25 days. These results proved that the multifunctional buffer layer engineering of AlO_x promoted the higher phase stability of wide-band gap perovskite films. The enhanced performance and stability of the wide-band gap PSCs based on the AlO_x layer should be responsible for the improved intrinsic stability of the perovskite film and restrained defect density at the perovskite/C₆₀ interface and perovskite grain boundary.

Furthermore, we fabricated two-terminal perovskite/silicon heterojunction TSCs using the modified perovskite to validate the applicability of the AlO_x layer in the perovskite/silicon TSCs. Figure 6a shows the device schematics of TSCs in this work, and the cross-section SEM image is shown in Figure S18. As shown in Figure 6b and S18, the best TSC after AlO_x modification exhibits a champion PCE of 28.50 % with a J_sc of 20.64 mA cm⁻², a V_oc of 1.819 V, and an FF of 75.92 %. The integrated J_sc from EQE spectra of the optimized tandem cell is 20.36 and 19.90 mA cm⁻² for the top and bottom sub-cells (Figure S20), respectively, which is within the accuracy confidence of the measurements. Detailed photovoltaic parameters are summarized in Table S4. Also, maximum power point tracking of the target TSC demonstrates stable power outputs with a stabilized PCE of 28.05 %, and there is no distinctive performance drop over 600 s (Figure 6c). The TSCs show good reproducibility with an average efficiency of 27.87 %, much better than the control devices (25.11 %). (Figure 6d and Figure S21). Moreover, the TSCs cooperating with the AlO_x layer show excellent light and thermal and moisture stability. As shown in Figure S22, the target TSCs sustain over 90 % of their initial PCEs over 800 hours under one-sun light soaking, continuous heating at 65 °C or relative humidity (RH) of 35±5%, respectively. The excellent efficiency and stability can be attributed to the higher intrinsic stability of AlO_x-based wide-band gap perovskite, suggesting the potential of the AlO_x layer for high-performance TSCs.

Conclusion

This work demonstrates a facile and effective strategy for precisely modulating the perovskite by incorporating AlO_x deposited by ALD on the top interface. Al³⁺ can not only infiltrate the bulk phase and interact with halide ions to suppress ion migration and phase separation, but also regulate the arrangement of energy levels and passivate defects on the perovskite surface and grain boundaries. Consequently, the ALD-AlO_x treatment can significantly improve the PCE to 21.80 % for wide band gap PSCs and 28.50 % for monolithic perovskite-silicon TSCs. More importantly, ALD-AlO_x exhibits a self-encapsulation effect through a dense interlayer, significantly improving the photothermal stability of PSCs. In addition, the resulting 1.55 eV PSC and module also achieved a PCE of 25.08 % (0.04 cm²) and 21.01 % (aperture area of 15.5 cm²), respectively, proving the universality of the strategy.

Conflict of interests

The authors declare no conflict of interest.