Abstract

Surface reflections and non-radiative recombinations create energy losses in perovskite solar cells (PSCs) by hindering the generation and extraction of carriers. These losses can reduce device efficiency in practical applications as the incident angle of sunlight varies throughout the day. Here we introduce a universal strategy to address this issue by coating glass substrates with highly distributed nanoplates of fluorine-doped tin oxide (NP-FTO). An electron-selective homojunction is then formed with a thin layer of SnO₂ deposited by atomic layer deposition covered with SnO₂ quantum dots. Systematic mechanistic studies reveal the exceptional ability of NP-FTO to harvest photons omnidirectionally and its beneficial influence on perovskite crystallization. These combined effects result in substantial improvements in the short-circuit current density, open-circuit voltage and fill factor of n–i–p PSCs under wide-angle incident light illumination. The best-performing PSC achieves a remarkable power conversion efficiency (PCE) of 26.4% (certified 25.9%) under AM1.5G illumination. The devices also show exceptional stability, retaining 95% of their initial PCE after 1,200 hours of light soaking under simulated solar intensity with maximum power point tracking. Moreover, the beneficial effects of NP-FTO are also applicable to 1.77 eV wide-bandgap PSCs with a p–i–n structure, enabling the fabrication of all-perovskite tandem solar cells with a best PCE of 28.2%.

Main

Metal halide perovskite solar cells (PSCs) have rapidly emerged as leading contenders in photovoltaic technologies, achieving power conversion efficiencies (PCEs) surpassing 26%, driven by advancements in device architecture and material engineering^{1,2,3,4,5,6,7,8,9}. However, most reported PCEs are measured under vertical simulated sunlight, overlooking severe energy losses across varying incident angles¹⁰. As sunlight angles shift throughout the day, PSCs experience substantial short-circuit current density (J_sc) and efficiency losses when light angles exceed 40°, limiting large-scale application^11,12. Minimizing both light energy loss across wide angles and non-radiative recombination is crucial for practicalimplementation.

Significant efforts have been dedicated to material engineering strategies, including bulk doping and interface passivation, for improving the crystallinity of perovskites and suppressing the non-radiative recombination^{13,14,15,16,17,18,19,20,21,22,23}. In contrast, essential influences of substrate structures on the crystallization process and resulting crystallinity of perovskites have been seldomly reported. In particular, the substrate structure significantly impacts light propagation into perovskite absorbers and the transfer behaviour of carriers at the buried interface^24,25. Textured substrates are used to effectively capture incident light, promoting light trapping by extending the optical path length and enhancing the J_sc (ref. ⁵). However, the mechanisms underlying the suppressed energy loss as well as the growth of perovskites on these textured structures remain subjects that require further in-depth investigation. Furthermore, a bilayer configuration of conformal charge-transporting layers has been developed to facilitate the charge transfer on textured substrates^5,26, while the selection of conformal transporting-layer materials also largely compacts on the light and carrier management. For instance, in n–i–p PSCs, SnO₂ has a refractive index (n) of about 1.9 (at 550 nm), offering superior optical compatibility with transparent conductive oxide (TCO) substrates including indium tin oxide (ITO) and fluorine-doped tin oxide (FTO)^27,28,29. In contrast, electron transport layers with higher refractive indices, such as TiO₂, introduce greater reflective losses at the interface. In addition, conduction band misalignment between FTO and TiO₂ can exacerbate charge extraction inefficiencies, further impairing device performance.

Here we use a tailored TCO substrate featuring an FTO-coated glass substrate with highly distributed nanoplates (NP-FTO). This configuration effectively reduces wide-angle light losses across a broad spectrum, outperforming normal TCO substrates, including standard ITO and standard FTO lacking such distinctive structures. More importantly, perovskites deposited on NP-FTO show enhanced crystallization and improved carrier mobilities. Thorough investigations into the underlying mechanisms responsible for these improvements have been conducted. Furthermore, through an all-SnO₂ electron-selective homojunction—combining atomic layer deposition (ALD) and a spin-coating technique—we minimize the light loss and non-radiative recombination, achieving a PCE of 26.4% (certified at 25.9%) for an n–i–p structure device. Moreover, NP-FTO also showed great potential in high-performance all-perovskite tandem solar cells, achieving a best PCE of 28.2%.

Light-loss suppression on NP-FTO substrates

To illustrate the benefits of the NP-FTO nanostructure, we conducted a comparative study against highly transparent TCOs including standard ITO and FTO lacking distinctive structures. Notably, NP-FTO showed an average grain size exceeding 300 nm, showcasing a morphology with a highly distributed V-concaves nanostructure (Supplementary Fig. 1), which is more pronounced when viewed at a tilt angle of 35° (Fig. 1a). Three-dimensional atomic force microscopy images (Supplementary Fig. 2) further visualized the depth distribution of the concave regions of NP-FTO, showing an obvious height difference of approximately 100 nm. These measurements confirmed the high roughness characteristic of NP-FTO and the photograph in Supplementary Fig. 3 shows the high haze characteristic of NP-FTO.

Fig. 1: Structure and reflection of NP-FTO.

Crystallization of perovskite on NP-FTO substrate

To better understand the influence of NP-FTO nanostructures on perovskite film formation, we conducted characterizations of the perovskite films prepared via a sequential deposition technique on the substrates pre-covered by an ETL³⁵.

Interestingly, the PbI₂ film deposited on the NP-FTO substrate showed larger grain sizes with a relatively more porous structure compared with those grown on ITO and FTO (Supplementary Fig. 12). The resulting perovskite films, following the introduction of formamidinium cation (FA⁺) and an annealing process, also showed larger grain sizes and enhanced crystallinity on the NP-FTO (Supplementary Figs. 12 and 13). While perovskites on both ITO and FTO substrates crystallized similarly, ITO showed slightly better light harvesting compared with FTO. Therefore, in the discussion below, ITO was selected as the representative TCO lacking a distinctive structure for comparison with NP-FTO’s influence on the perovskite crystallinity mechanism.

We first measured the contact angles of the PbI₂ solution drops on ITO/ETL and NP-FTO/ETL (Supplementary Fig. 14). Notably, the contact angle of the PbI₂ precursor on ITO/ETL (15.7°) exceeded that on NP-FTO/ETL (0°), suggesting better wetting on the nanostructured NP-FTO surface. In addition, the highly distributed V-concaves are expected to provide increased depth for precursor permeation, thereby potentially slowing down the evaporation of dimethylformamide or dimethyl sulfoxide solvent and facilitating the growth of the PbI₂ film during both the spin-coating and annealing processes³⁶. Building on these insights, we performed in situ X-ray diffraction (XRD) measurements to determine the crystallization dynamics of perovskites during the incorporation of organic ammonium salts into PbI₂ layers (Fig. 2a,b). A distinct diffraction peak assigned to the (001) plane of α-phase FAPbI₃ was detected at 14° on both ITO-supported and NP-FTO-supported samples. Notably, an additional intermediate phase of PbI₂(DMSO)₂ was observed at about 10.3° during the deposition on ITO³⁷, indicating gradual FA⁺ penetration into the PbI₂ skeleton. In contrast, a single but more pronounced peak of α-phase FAPbI₃ was detected on the NP-FTO-supported film, confirming the facilitated infiltration of FA⁺ and favourable formation of black-phase perovskites. A schematic of the crystallization mechanism during these processes is shown in Supplementary Fig. 15, illustrating the beneficial effect of the mesostructure of NP-FTO on improving the crystallinity of perovskites.

Fig. 2: Crystallization properties and QFLS on NP-FTO.

Suppression of non-radiative recombination in NP-FTO-based devices

Interface engineering through conformal deposition is also essential for optimizing NP-FTO-based devices. Upon examining the SnO₂ films on the substrates via scanning electron microscopy (SEM) analysis (Supplementary Fig. 20), it was found that the ITO surface was fully covered by spun-cast SnO₂ nanoparticles, but NP-FTO remained partially exposed, potentially causing severe non-radiative recombination. To address this issue, a 4-nm-thick SnO₂ thin film was first deposited using ALD (denoted as ALD-SnO₂ hereafter), forming a conformal layer on the surface of the NP-FTO substrate. This was followed by the deposition of a spun-cast SnO₂ quantum film (hereafter denoted as SC-SnO₂), creating a bilayer ETL that effectively covered the NP-FTO surface (Supplementary Fig. 20c).

Cross-sectional transmission electron microscopy images, along with energy-dispersive X-ray spectroscopy element mapping, clearly indicated comprehensive coverage of the spire and ridge of the NP-FTO nanostructure after the incorporation of ALD-SnO₂ (Supplementary Fig. 21). To further explore the impact of ALD-SnO₂ on the energy-level alignment and carrier extraction, we conducted ultraviolet photoelectron spectroscopy measurements (Supplementary Fig. 22). The alignment diagram constructed from these measurements revealed a smooth energy-level ladder that facilitates electron extraction after introducing ALD-SnO₂ (Supplementary Fig. 23). Furthermore, we used absolute photoluminescence intensity and photoluminescence quantum yield measurements to create the depth-averaged maps of quasi-Fermi-level splitting (QFLS) for perovskite films deposited on ITO/ALD-SnO₂/SC-SnO₂ and NP-FTO/ALD-SnO₂/SC-SnO₂ (refs. ^40,41; Supplementary Note 2). Notably, the depth-averaged QFLS for the perovskite film on ITO was about 1.205 eV, whereas it was near 1.220 eV for NP-FTO (Fig. 2g–i), suggesting its effective suppression of non-radiative recombination.

Performance and stability of n–i–p devices based on NP-FTO

Building on the positive effects of NP-FTO with an ALD-SnO₂ layer, we fabricated n–i–p structured devices (Fig. 3a). Remarkably, the NP-FTO device achieved a best PCE of 26.40% in a reverse scan and 26.25% in a forward scan (Fig. 3b), significantly outperforming ITO- and FTO-based devices (Supplementary Fig. 25). The certified PCE for the NP-FTO device reached 25.88% (Supplementary Fig. 26). External quantum efficiency (EQE) characterizations on both ITO- and NP-FTO-based devices provided consistent J_sc results with the current density–voltage (J–V) curves (Fig. 3c). The best NP-FTO-based device also maintained a stabilized PCE of 26.29% after maximum-power-point tracking (MPPT) for 600 s (Fig. 3d). Substitution of ITO with NP-FTO also resulted in statistical enhancements in J_sc (about 1 mA cm⁻²), open-circuit voltage (V_oc, about 0.015 V) and about 2% for fill factor (FF) as the results shown in Fig. 3e. The effects of film thickness and annealing temperature of ALD-SnO₂ on the photovoltaic performance are also investigated in Supplementary Figs. 27 and 28 and Supplementary Tables 1 and 2. To further explore the light utilization across a wide range of incident angles, we tilted the device in an anticlockwise direction from 0° to 45°. With increasing light incident angle, the device’s PCE decreased (Fig. 3f,g and Supplementary Table 3). Notably, the reduction of J_sc in the ITO-based device was more pronounced, exceeding 5% as the incident angle increased from 0° to 45°, whereas the NP-FTO device showed almost no attenuation (Fig. 3h). This highlighted NP-FTO’s superior light utilization and reduced sensitivity to light incident angles.

Fig. 3: Photovoltaic characteristics.

Universality of NP-FTO on wide-bandgap and all-perovskite tandem solar cells

All-perovskite tandem solar cells have shown the potential to surpass the Shockley–Queisser limit of single-junction PSCs^{44,45,46,47,48}. To demonstrate the universality of NP-FTO, here we fabricated all-perovskite tandem solar cells comprising a 1.77 eV wide-bandgap (WBG) subcell and a 1.25 eV narrow-bandgap (NBG) subcell. A cross-sectional SEM image of the tandem revealed that the WBG perovskite film completely covered the hole transport layer on the substrate surface (Fig. 4a), resulting in negligible substrate-induced differences in the NBG perovskite film deposited atop the WBG sublayer, regardless of the substrate used (Supplementary Fig. 35). In addition, as the WBG perovskite film has a critical role in tandem solar cells performance, we then put emphasis on the effects of NP-FTO on this layer. Uniquely, our approach uniquely investigated the combined effects of reflection suppression and non-radiative recombination reduction enabled by NP-FTO.

Fig. 4: Universality of NP-FTO on all-perovskite tandem solar cells.

Methods

Materials

SnO₂ colloid precursor (tin(iv) oxide, 15% in H₂O colloidal dispersion) was purchased from Alfa Aesar. N,N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), isopropanol (IPA), chlorobenzene, acetonitrile, 4-tert-butylpyridine (t-BP), bis(trifluoromethane)sulfonimide lithium salt (Li-TFSI), lead thiocyanate (Pb(SCN)₂), caesium chloride (CsCl), potassium iodide, diethyl ether and methylammonium chloride (MACl) were purchased from Sigma-Aldrich. Lead iodide (PbI₂, >98%) and [4-(3,6-dimethyl-9H-carbazol-9-yl)butyl]phosphonic acid (Me-4PACz) were purchased from TCI. Formamidinium iodide (FAI), caesium iodide (CsI), lead bromide (PbBr₂), NiO_x, methylammonium iodide (MAI), SnI₂ and 2,2′,7,7′-tetrakis(N,N-di-p-methoxy-phenylamine)-9,9′-spirobifluorene (spiro-OMeTAD, 99%) were purchased from Advanced Election Technology. Ethane-1,2-diammonium iodide (EDAI₂), bathocuproine (BCP), 1,3-propyldiammonium diiodide (PDAI₂), n-butylammonium bromide (BABr) and poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA) were purchased from Xi’an Polymer Light Technology. Poly(3,4-ethylenedioxythiophene)/poly(styrenesulfonate) (PEDOT:PSS) aqueous solution (Al 4083) was purchased from Heraeus Clevios. Hydrogen peroxide aqueous solution (H₂O₂, 30%) was purchased from Sinopharm Chemical Reagent. C₆₀ and copper were purchased from Xi’an Yuri Solar. NP-FTO (9 Ω sq⁻¹) substrates were purchased from Asahi. ITO (9 Ω sq⁻¹) and standard FTO (9 Ω sq⁻¹) substrates were purchased from Advanced Election Technology and Pilkington.

Solution preparation

Single-junction PSCs (n–i–p)

SnO₂ colloid solutions (15 wt%) were diluted with H₂O₂ and deionized water (1:1:4, v:v:v). PbI₂ solutions (1.5 M) were prepared by dissolving 691.5 mg PbI₂ powder, 1.5 mg Pb(SCN)₂, 2.5 mg potassium iodide and 5 mg CsCl in 1 ml DMF and DMSO (9:1, v:v) and stirred at 70 °C for 4 h. Organic amine salt solutions were prepared by dissolving 90 mg FAI and 12–16 mg MACl in 1 ml IPA and stirred at room temperature for 4 h. The BABr solutions were prepared by dissolving 3.0 mg BABr in 1 ml IPA. Spiro-OMeTAD solutions were prepared by dissolving 72.3 mg spiro-OMeTAD, 28.8 μl t-BP and 17.5 μl Li-TFSI solution (520 mg ml⁻¹, in acetonitrile) in 1 ml chlorobenzene. PTAA solutions were prepared by dissolving 24 mg PTAA, 12 μl t-BP and 5 μl Li-TFSI solution (260 mg ml⁻¹, in acetonitrile) in 1 ml chlorobenzene.

All-perovskite tandem PSCs (p–i–n)

WBG subcells

NiO_x solutions were prepared by dissolving 5 mg NiO_x in 1 ml deionized water. Me-4PACz solutions were prepared by dissolving 0.5 mg Me-4PACz in 1 ml ethanol solution. FA_0.8Cs_0.2PbI_1.8Br_1.2 (1.1 mol) perovskite precursors were prepared by dissolving FAI, CsI, PbI₂, PbBr₂ and Pb(SCN)₂ in a mixed solvent of DMF and DMSO with a volume ratio of 3:1. PDAI₂ solutions were prepared by dissolving 2.0 mg PDAI₂ in 1 ml IPA⁴⁹.

NBG subcells

PEDOT:PSS solutions were diluted in IPA (1:2, v:v). FA_0.7MA_0.3Pb_0.5Sn_0.5I₃ (1.8 mol) perovskite precursor solutions were prepared by dissolving FAI, MAI, PbI₂, SnI₂, SnF₂ and Pb(SCN)₂ in a mixed solvent of DMF and DMSO with a volume ratio of 3:1. EDAI₂ solutions were prepared by dissolving 0.5 mg EDAI₂ in 1 ml IPA.

Device fabrication

PSCs with n–i–p structure

Substrates were sonicated in deionized water, acetone and ethanol in sequence. Then the substrates were dried using a nitrogen stream and turned to a thermal ALD process without an ultraviolet ozone process. For the ALD process, tetrakis(dimethylamino)tin(iv) was used as the Sn precursor and preheated at 95 °C in a stainless-steel container and water was used as the oxidant, which was preheated at 45 °C. The deposition process was conducted to form 4 nm (24 cycles) of SnO₂ film at 90 °C and then annealed at 150 °C for 30 min in ambient conditions. Next, SnO₂ colloid solutions were spun-cast onto the substrates at 4,000 rpm for 30 s and annealed at 150 °C for 30 min in ambient air. After cooling, PbI₂ solutions were spin-coated at 1,500 rpm for 30 s and annealed at 70 °C for 1 min in a N₂-filled glovebox. Then the organic amine salt solutions were spun-cast onto the PbI₂ films, at 2,000 rpm for 30 s and annealed in the air (30–40% relative humidity) at 145 °C for 13 min. For BABr treatment, the BABr solutions were spun-cast onto the perovskite film surfaces at 3,000 rpm for 30 s and annealed at 100 °C for 1 min to dry residual IPA. Hole extraction layers were prepared by spin-casting the spiro-OMeTAD solutions at 3,000 rpm for 20 s. For the optional stability devices, the PTAA solution was spun-cast at 2,000 rpm for 40 s. Finally, 70 nm of gold was thermally evaporated to complete the whole device. A 0.070025 cm² non-reflective mask was used to define the devices’ active area and another mask with a mask area of 0.0692 cm² was used to certify the devices’ efficiency.

WBG PSCs (p–i–n)

NiO_x solutions were spin-coated on the substrates at 3,000 rpm for 30 s, followed by annealing at 130°C for 30 min and then transferred to an N₂-filled glovebox. Me-4PACz solutions were spin-coated on NiO_x at 4,000 rpm for 30 s, followed by annealing at 100 °C for 10 min. WBG perovskite solutions were spin-coated at 4,000 rpm for 30 s. At the last tenth of a second of this process, 500 μl diethyl ether was dropped as antisolvent. After that, the perovskite films were annealed at 70 °C for 2 min and 100 °C for 8 min. Then PDAI₂ solutions were deposited on the perovskite films at 4,000 rpm for 30 s and annealed at 100 °C for 5 min. Finally, C₆₀ (20 nm), BCP (7 nm) and copper (80 nm) were thermally evaporated to complete the devices.

All-perovskite tandem

The WBG subcells fabrication were completed as described above until the deposition of ALD-SnO₂ (20 nm). After that, 1 nm gold was thermally evaporated to serve as the recombination layer. Next, PEDOT:PSS solutions were spin-coated onto the WBG subcells at 4,000 rpm for 30 s and annealed at 120 °C for 10 min. Substrates were then transferred to a N₂-filled glovebox. NBG perovskite precursor solutions were spin-coated onto substrates at 1,000 rpm for 10 s, followed by 3,800 rpm for 45 s. At 20 s before the end of the second step, 400 µl chlorobenzene was dropped onto the substrate, which was then annealed at 100 °C for 10 min. Then EDAI₂ solutions were deposited at 4,000 rpm for 25 s, followed by annealing at 100 °C for 5 min. Finally, C₆₀ (20 nm), BCP (7 nm) and copper (80 nm) were thermally evaporated to complete the devices.

Photovoltaic characterization

The J–V characteristics of the PSCs were recorded under a standard AM1.5G solar simulator (class AAA, 450 W xenon lamp, Newport 94043A) using a Keithley 2400 source meter, and the light intensity was calibrated using a KG-5 filtered silicon diode. The devices were measured in reverse-voltage scans. The corresponding EQEs were measured using a QE/IPCE system (Enli Technology). The electroluminescence efficiencies of the devices were determined by measuring the emitted photons of the devices in all directions through an integrated sphere using a calibrated spectrometer (QE Pro, Ocean Optics), under a constant current density provided by a Keithley 2400 source measure unit. The MPPT measurements were conducted using a white light-emitting diode source (wavelength from 410 nm to 850 nm) with an AM1.5 filter and an irradiation intensity of 100 mW cm⁻² in an N₂ atmosphere using a customized photovoltaic test system integrated into a glovebox (PLVT-G8001X-16B). The device temperature was measured to be about 55 °C using a thermocouple.

Film characterization

The transmission electron microscopy images and corresponding energy-dispersive X-ray spectroscopy element mapping were recorded by a high-angle annular dark-field transmission electron microscope (JEOL JEM-F200). The SEM images were obtained by using a field-emission scanning electron microscopy (Zeiss GeminiSEM 500) and three-dimensional atomic force microscopy images of perovskite films were measured using a Bruker NanoScope 8 atomic force microscope. TRPL was performed using a Delta Flex fluorescence spectrum spectroscopy (HORIBA). For TRMC measurements, the samples were pumped with a tunable optical parametric oscillator (OPO) laser, which had a pulse width of approximately 5 ns and a repetition rate of 10 Hz. A FieldFox Handheld Microwave Analyzer (Keysight, N9915A) was used as a microwave source and detector. ARR measurements were carried out through an angle-resolved micro spectroscopy system, which was obtained from Shanghai Fuxiang Optics. The incident light was generated from the system at all angles and then transformed into a tiny facula by Fourier transform. The reflection angle was the collecting angle. As the collecting angle tended to 0, there was the strongest reflection. The reflectance results were obtained by using an integrating sphere with incident light in a vertical direction. XRD patterns of perovskite films were recorded using a Rigaku smartlab XRD instrument with Cu Kα radiation under operating conditions of 40 kV and 44 mA. Ultraviolet photoelectron spectroscopy measurements of SnO₂ films were performed using an X-ray photoelectron spectroscopy/ultraviolet photoelectron spectroscopy system (ESCLAB 250Xi, Thermo Scientific) with He Iα radiation.