Introduction

Organic–inorganic hybrid perovskite solar cells (PSCs) have recently achieved a power conversion efficiency (PCE) exceeding 26% [1]. This achievement is due to the high absorption coefficient, high charge mobility, and low exciton binding energy inherent to perovskite materials [2,3,4,5]. A notable advantage of these materials is their exceptional mechanical flexibility, which stems from their hybrid organic–inorganic structure [6, 7]. Consequently, flexible PSCs have been extensively investigated, demonstrating high mechanical stability and performance [8,9,10].

Fabricating flexible PSCs typically involves constructing each layer from a polymer substrate. To provide the flexibility, polymer substrates such as polyethylene naphthalate (PEN), known for their suitable optical and mechanical properties, are commonly used [9, 11, 12]. However, these polymer substrates exhibit higher moisture and oxygen permeability than glass substrates [13,14,15]. Considering that perovskite materials are easily degraded by moisture and oxygen, polymer substrates are less suitable for long-term applications [16, 17]. Ultrathin glass (UTG) substrates present a viable alternative to polymer substrates. With a thickness of less than 100 μm, UTG maintains adequate flexibility while offering the advantageous properties of glass, including a moisture permeability over 100,000 times lower than that of polymer substrates [14, 18, 19]. Consequently, UTG-based flexible PSCs are expected to achieve long-term moisture stability.

Another key consideration in fabricating flexible PSCs is the selection of an efficient electron transport layer (ETL). The ETL is a crucial component in PSCs, significantly affecting their performance and stability [20, 21]. Positioned between the transparent electrode and the perovskite layer, the ETL efficiently extracts photogenerated electrons from the perovskite layer to the electrode. For effective hole blocking, it is essential for the ETL to completely cover the bottom electrode. Additionally, the ETL should exhibit favorable energy level alignment between the perovskite and the transparent electrode to ensure efficient electron extraction. One strategy to achieve this is the implementation of a bilayer ETL with a favorable energy level design. Recent studies have indicated that bilayer ETLs can provide higher PCE and enhanced stability than monolayer ETLs [9, 11, 22].

In this study, we designed a SnO2/TiO2 bilayer for the ETL to fabricate flexible PSCs. Initially, we synthesized colloidal TiO2 nanocrystals (NCs), which were capped with oleic acid (OA) to ensure dispersibility in hexane, a non-polar solvent. To prepare the bilayer ETL, TiO2 dispersion was spin-coated onto a SnO2 film, and the organic ligands were removed via UV irradiation. We confirmed that this TiO2 layer possesses favorable energy level alignment between the SnO2 and perovskite layers, resulting in efficient charge extraction. The bilayer ETL-based PSCs exhibited a PCE of 23.5%, an improvement over the 22.7% PCE of PSCs with a SnO2 monolayer ETL. Moreover, PSCs using UTG substrates showed significant differences in PCE, with bilayer and monolayer ETLs achieving 15.2% and 8.3%, respectively. We demonstrated that the performance degradation due to UTG roughness is mitigated by employing the bilayer ETL. Furthermore, we highlighted the effectiveness of the UTG substrate in preventing moisture penetration by comparing UTG-based PSCs with PEN-based PSCs.

Experimental

Chemicals

Tin(IV) oxide (SnO2 NCs, 15% in H2O colloidal dispersion), formamidine acetate salt (99%), and lead(II) iodide (PbI2, 99.999%) were purchased from Alfa Aesar. Oleic acid (OA, 90%), titanium(IV) isopropoxide (TTIP, 97%), n-hexane (anhydrous, 95%), hydroiodic acid (HI, 57 wt% in water), 2-methoxyethanol (2ME, anhydrous, 99.8%), 2-propanol (IPA, anhydrous, 99.5%), chlorobenzene (CB, anhydrous, 99.8%), N,N-dimethylformamide (DMF, anhydrous, 99.8%), dimethyl sulfoxide (DMSO, > 99.5%), methylammonium chloride (MACl, 98%), acetonitrile (ACN, anhydrous, 99.8%), and 4-tert-butylpyridine (tBP, 98%) were purchased from Sigma–Aldrich. Phenethylammonium iodide (PEAI) was acquired from Greatcell Solar. Ethyl alcohol (EtOH, pure) and diethyl ether (extra pure grade) were purchased from Duksan. 2,2′,7,7′-Tetrakis(N,N-di-p-methoxyphenylamine)-9,9′-spirobifluorene (Spiro-OMeTAD) and Tris(2-(1H-pyrazol-1-yl)-4-tert-butylpyridine)-cobalt(III) Tris(bis(trifluoromethylsulfonyl)imide) (FK209) were obtained from Lumtec. Ultrathin glass (UTG, 50 µm) based tin-doped indium oxide (ITO) substrates were purchased from Quantum Plasma, Korea.

Synthesis of Colloidal TiO2 Nanocrystals

TiO2 NCs were synthesized through a modified non-hydrolytic sol–gel reaction method, as reported previously [23]. The synthesis began by combining 88 mmol of OA and 30 mmol of TTIP in a three-neck round-bottom flask under a continuous argon (Ar) gas flow. This mixture was heated to 270 °C for 2 h with constant stirring. During the formation of TiO2 NCs, the reaction mixture's color transitioned from transparent yellow to white. The precipitates were washed with excess ethanol and purified by centrifugation at 3000 rpm for 30 min. The final product, dissolved in n-hexane at 25 mg·mL−1, was stored under an N2 gas within a glove box.

Synthesis of the Perovskite Precursor

Formamidinium iodide (FAI) and formamidinium lead triiodide (FAPbI3) powders were synthesized using previously established methods [21]. To synthesize FAI powder, formamidine acetate salt (25 g) was mixed with HI (50 mL) in a 500 mL round-bottom flask and stirred vigorously. The solvent was evaporated under vacuum at 80 °C for 1 h, yielding a light-yellow powder. This powder was dissolved in ethanol and precipitated with diethyl ether, repeated three times until a pure white powder was obtained. For further purification, the FAI powder was recrystallized using an ethanol and diethyl ether mixture (1:1 v/v) in a refrigerator. The final white powder was filtrated through a glass filter and dried at 60 °C for 24 h. For synthesizing FAPbI3 powder, the prepared FAI (0.8 M) was mixed with PbI2 in a 1:1 molar ratio in 2ME with vigorous stirring. The resulting yellow solution was heated and stirred vigorously at 120 °C, then subjected to recrystallization using the retrograde method. The final black powder was filtered through a glass filter and baked at 150 °C for 1 h to obtain the desired FAPbI3 powder.

Fabrication of Perovskite Solar Cells

Patterned ITO substrates were sequentially cleaned by sonication in detergent water, deionized water, and IPA for 10 min each. To produce SnO2 ETLs, the SnO2 NCs were diluted with deionized water (SnO2 NCs:H2O = 1:6). This solution was dropped onto UV-ozone-treated transparent ITO substrates and spin-coated at 3000 rpm for 30 s, followed by annealing at 150 °C for 30 min. To prepare the SnO2/TiO2 bilayer electron transport layer, a highly dispersed solution of TiO2 NCs in n-hexane (25 mg·mL−1) was dropped onto a transparent conducting oxide (TCO) substrate and immediately spin-coated at 5000 rpm for 30 s. The TiO2 NC-coated substrate was then treated with UV irradiation using a UV cure machine (JHCI-051B, JECO) equipped with a mercury lamp (peak wavelength: 365 nm). For perovskite layer deposition, the UV-ozone-treated substrate was spin-coated with the perovskite precursor solution at 8000 rpm for 50 s. After 10 s of spin coating, 1 mL of diethyl ether was dropped onto the spinning film. The deposited film was then annealed at 150 °C for 15 min and 100 °C for 1 h on a hotplate. The perovskite precursor solution was prepared by dissolving 778 mg of synthesized FAPbI3 powder and 23 mg of MACl in 0.5 mL of DMF/DMSO (4:1 v/v). After cooling, the perovskite films were coated with PEAI (3 mg·mL−1 in IPA) at 3000 rpm for 30 s. Spiro-OMeTAD, the hole transport layer (HTL) in the PSCs, was spin-coated onto the films at 4000 rpm for 30 s. The Spiro-OMeTAD solution was prepared by doping 90 mg/mL Spiro-OMeTAD in CB with 39 µL of tBP, 23 µL of Li-TFSI (520 mg·mL−1 in ACN), and 5 µL of FK209 (180 mg·mL−1 in ACN). Finally, an Au electrode with a thickness of 80 nm was deposited using a thermal evaporator.

Characterization

The crystal structure of the TiO2 NCs was characterized using a high-resolution X-ray diffractometer (XRD, D8 ADVANCE, BRUKER) with Cu-Kα radiation (λ = 1.541 Å). The morphology of SnO2 and TiO2 NCs was analyzed with transmission electron microscopy (TEM, JEM2100F, JEOL), while their films were examined using scanning electron microscopy (FE-SEM, Verios G4 UC, FEI) and atomic force microscopy (AFM, XE-100, Park Systems). The optical properties of the thin films were investigated using a UV–Vis spectrometer (Shimadzu, UV-2600i). The absorption coefficient (α) for direct bandgap semiconductors near the band edge was calculated using the formula αhν = A(hν–Eg)2, where A denotes the proportionality constant, hν denotes the photon energy, and Eg denotes the optical bandgap. Specifically, Eg was estimated by extrapolating the tangent line to the plotted curve. ATR–FTIR spectroscopy (Nicolet iS50, Thermo Fisher Scientific) was employed to measure ligand removal after UV treatment. Steady-state PL and TRPL spectra were obtained using a Fluoromax-4 (iHR320, HORIBA Scientific) spectrometer, with an excitation wavelength of 463 nm. TRPL measurements were conducted to determine the PL maximum (~ 800 nm) of the perovskite film. Under dry conditions (20 ± 5 °C, below 15% RH), the current density–voltage (J–V) curve and MPP tracking characteristics of the devices were measured using a Keithley 2400 source meter under a Xenon-lamp-based solar simulator (Newport 91160 s, AAA class). A Si solar cell calibrated by the National Renewable Energy Laboratory (NREL) was used to adjust the light intensity to AM 1.5G 1 sun conditions (100 mW·cm−2). The voltage sweep rate was 20 mV with a delay of 20 ms for the reverse (forward) sweeping direction from 1.3 V (− 0.1 V) to − 0.1 V (1.3 V). The devices were measured with a SUS304 aperture with an area of 0.09 cm2. The incident photon-to-current conversion efficiency (IPCE) was characterized using an IPCE measurement system (PV Measurement, Inc.).

Results and Discussion

SnO2 NCs are commonly used as the ETL in PSCs [24]. However, achieving perfect hole-blocking ability with SnO2 ETL alone is challenging because it is difficult to fully cover the surface while maintaining a thin layer. As the thickness of the SnO2 ETL increases, the PCE tends to decrease due to heightened resistance [9]. A facile strategy to address this issue is the sequential deposition of different ETLs to create a bilayer ETL. When forming a bilayer ETL, the lower layer should possess a higher conduction band minimum (CBM) to enhance electron extraction [9, 11, 22]. Anatase TiO2, with a slightly higher CBM than SnO2, is an appropriate choice [25, 26]. Consequently, we developed a bilayer ETL with a TiO2 layer on top of a SnO2 layer.

To minimize the impact on the SnO2 layer, which can dissolve in polar solvents, we synthesized TiO2 NCs dispersed in a non-polar solvent. The transmission electron microscopy (TEM) image of the synthesized TiO2 NCs is shown in Figure S1, confirming the rod-shaped nanocrystals as previously reported [23]. These TiO2 NCs are capped with a long-chain organic ligand (OA) to facilitate dispersion. However, these surface stabilizers were removed by UV irradiation in the thin-film state, as they would impede charge transport [10, 23]. The removal of organic ligands by UV irradiation was confirmed by ATR-FTIR analysis (Fig. S2). The UV-sintered TiO2 film does not exist strong vibration peaks at 2925 cm–1 and 2855 cm–1, corresponding to the asymmetric and symmetric CH2 stretching vibrations of OA. Other vibration peaks of the organic ligands, such as COO– (1525 cm–1, asymmetric) and CH2 (1440 cm–1, bending), also disappear after the UV sintering.

X-ray diffraction (XRD) patterns of the TiO2 film showed that all reflection peaks correspond to the anatase TiO2 crystal structure (PDF #21-1272) (Fig. S3). The SnO2/TiO2 bilayer film was prepared using the UV-sintering method for the ETL, as depicted in the scheme of Fig. 1a. We initially compared the transmittance of both SnO2 and SnO2/TiO2 films. The UV–Vis transmittance of each layer on glass/ITO substrates is shown in Fig. 1b. The SnO2/TiO2 film demonstrated a slightly higher transmittance than the SnO2 film at 350–500 nm, indicating that more near-UV light could reach the light-absorbing layer. Figure 1c illustrates the band alignment for the designed ETLs, with this band diagram estimated from ultraviolet photoelectron spectroscopy (UPS) and Tauc plots in Fig. S4. The estimated CBM of SnO2, TiO2, and perovskite are − 4.30 eV, − 4.20 eV, and − 3.96 eV, respectively. The results indicate that the CBMs of TiO2 and SnO2 are arranged in a cascade, providing favorable energy levels for charge extraction [22].

Fig. 1
figure 1

a Schematic illustration of the UV sintering process for SnO2/TiO2 bilayer formation. b Transmittance spectra of SnO2 and SnO2/TiO2 films on glass/ITO. c Energy level diagrams of SnO2, TiO2, and perovskite calculated from the Tauc plot and UPS spectrum. d Steady-state PL spectra of the perovskite films deposited on SnO2 and SnO2/TiO2 films. e TRPL spectra with an excitation wavelength of 463 nm for perovskite films deposited on SnO2 and SnO2/TiO2 films

Full size image

The steady-state photoluminescence (PL) spectra and time-resolved PL (TRPL) decay was measured to investigate the charge transfer ability at the interface of each ETL and the perovskite. Figure 1d illustrates that devices utilizing the bilayer ETL exhibit more effective PL quenching than those employing only the SnO2 layer, indicating that charge extraction from perovskite to the bilayer ETL is more efficient than with only a SnO2 layer [27,28,29]. Figure 1e shows the PL decay transients of perovskite coated on each ETL, from which the corresponding PL lifetimes and amplitudes were obtained by fitting with a bi-exponential decay function, as listed in Table S1. The fast decay component (\({\tau }_{1}\)) originates from the quenching of charge carriers at the interface, while the slow decay component (\({\tau }_{2}\)) is attributed to the radiative recombination of free charge carriers due to traps in the absorption layer [30]. Compared to the perovskite-only film, the samples using ETL exhibited a shorter \({\tau }_{1}\) with an increasing amplitude of the fast decay (\({A}_{1}\)), indicating that charge transfer from the perovskite film to the ETL became dominant. The average lifetime for the SnO2/perovskite sample was 251.88 ns, which reduced to 172.13 ns for the bilayer ETL. Additionally,\({\tau }_{1}\) decreased from 164.17 to 71.65 ns in the bilayer ETL sample, with an increasing \({A}_{1}\). This implies that the bilayer ETL is more efficient in extracting charge from the perovskite.

To compare the thickness and uniformity of each SnO2 film, we used scanning electron microscopy (SEM) to measure each ETL-based PSC (device structure: ITO/ETL/perovskite/Spiro-OMeTAD/Au). These images show thicknesses of 20–30 nm for the SnO2 layer and 60–80 nm for the bilayer (Fig. 2a and b). Additionally, they indicate that the bilayer ETL is deposited uniformly. We also compared the morphology of perovskite films deposited on each ETL layer, revealing that the underlying layer does not significantly affect the perovskite crystal grains (Fig. S5).

Fig. 2
figure 2

Cross-sectional SEM image of a SnO2 and b SnO2/TiO2 bilayer ETL-based PSCs. c J–V characteristics of conventional glass/ITO substrate-based PSCs using SnO2 and SnO2/TiO2 bilayer ETLs. d J–V curves of UTG/ITO substrate-based PSCs with various ETLs

Full size image

Figure 2c shows the J–V curves of each ETL-based solar cell, with their photovoltaic performances summarized in Table S2. The bilayer ETL-based device exhibited a higher PCE compared to the SnO2-based devices, achieving a PCE of 23.47% (open-circuit voltage (VOC) = 1.19 V, short-circuit current density (JSC) = 23.43 mA·cm−2, fill factor (FF) = 84.07%). In contrast, the SnO2 ETL-based device had a PCE of 22.66% (VOC = 1.19 V, JSC = 22.87 mA·cm−2, FF = 83.04%). We measured the IPCE spectra to assess the charge collection efficiency of each ETL on the photocurrent of PSCs. The J–V characteristics measured in both reverse and forward directions of the devices show negligible differences (Fig. S6a). Figure S6b presents the IPCE spectra of solar cells with different ETLs, which align well with the values obtained from the J–V curves in Fig. 2c. The solar cells based on the bilayer ETL exhibited higher IPCE values with a strong spectral response from 350 to 500 nm. Photons with shorter wavelengths have a shallower penetration depth and are typically absorbed near the front region of the perovskite layer [10, 26]. Therefore, the observed increase in IPCE values for shorter wavelengths implies a more efficient electron extraction at the interface between the SnO2/TiO2 bilayer and the perovskite, as well as improved optical transparency of the bilayer ETL.

For flexibility, we fabricated a device with the same structure on an ITO-deposited ultrathin glass (UTG) substrate. The thickness of conventional glass/ITO is 1100 μm, whereas the UTG/ITO used is 50 μm. It is noteworthy that performance differences between each ETL-based PSC are more pronounced when using UTG substrates. In Fig. 2d, we compared the J–V curves of each ETL-based solar cell employing UTG substrates. The bilayer ETL demonstrated a PCE of 15.2%, while the SnO2 ETL-based device exhibited a significantly lower PCE of 8.3%.

To understand this performance difference, we compared the properties of conventional ITO substrates and UTG substrates. The UV–Vis transmittance of each substrate revealed slightly lower transmittance for UTG substrates (Fig. S7). This might explain the lower JSC value of UTG-based devices. Furthermore, we compared the roughness of each substrate using atomic force microscopy (AFM). As shown in Fig. 3a, the roughness of the UTG substrate (RMS of 81.0 nm) is significantly higher than that of the conventional substrate (RMS of 1.4 nm). Additionally, we compared films deposited with a SnO2 layer and a SnO2/TiO2 bilayer on the UTG substrate (Fig. 3b). The SnO2 film (RMS of 2.6 nm) exhibited higher roughness than the bilayer film (RMS of 1.0 nm). Moreover, images of the SnO2 film revealed that SnO2 alone could not sufficiently cover the high roughness of the UTG surface, exposing the ITO layer (Fig. 3c). These results indicate that the bilayer ETL not only improves the performance of PSCs but also provides sufficient coverage even for films with significant roughness.

Fig. 3
figure 3

a AFM images of conventional glass/ITO and UTG/ITO substrates. b AFM images of SnO2 layer and SnO2/TiO2 bilayer coated on a glass/ITO substrate. c Schematic representations of PSCs using SnO2 and SnO2/TiO2 ETLs

Full size image

We fabricated devices using UTG/ITO and PEN/ITO substrates to assess their moisture permeability. Water was applied to the back side of the device in direct contact with the substrate, and the PCE was monitored for over 1000 min. The initial J–V curves of the devices used for comparison are depicted in Fig. S8. The PEN-based device achieved a PCE of 18.0%, slightly higher performance than the UTG-based device, which had a PCE of 15.2%. After 1000 min, the PEN-based device showed a 71% reduction in its initial PCE (Fig. 4a). Conversely, the UTG-based device exhibited a relatively low moisture permeability, with only a 5% reduction in initial PCE. The optical images in Fig. 4b visually illustrate the state of the devices, revealing that the perovskite layer in the PEN-based device dissolved, whereas the UTG-based device did not. Given that water was applied only to the backsides of the devices, this indicates that moisture penetration from the substrate influenced the outcome. Thus, UTG substrates are recommended to fabricate flexible perovskite solar cells that are moisture-resistant.

Fig. 4
figure 4

a Normalized PCE obtained from moisture permeability tests of PEN/ITO, and UTG/ITO substrate-based PSCs. b Optical images of each PSC before and after the moisture permeability test

Full size image

Conclusions

In this study, we demonstrated that a strategically designed SnO2/TiO2 bilayer is an efficient ETL for PSCs, enhancing charge carrier extraction from the perovskite to the transparent electrode by modifying the ETL's energy levels. This bilayer ETL proved more effective on substrates with high roughness, such as UTG substrates, due to its superior coverage compared to a monolayer ETL, resulting in better hole-blocking properties. Consequently, the PCE of bilayer ETL-based PSCs improved to 15.2%, compared to 8.3% for monolayer ETL-based PSCs. This demonstrates the bilayer ETL's efficiency even on substrates with significant roughness.

Furthermore, we emphasized the low moisture permeability of UTG substrates compared to conventional polymer substrates. Moisture permeability tests indicated that devices based on polymer substrates experience a sharp decline in PCE, while those based on UTG substrates maintain their PCE. This suggests the potential of UTG substrates for the long-term use of flexible PSCs.