Introduction

Owing to the singular properties, including tuned band gap, small exciton energy, excellent bipolar carrier transport, long charge diffusion length, and amazingly high tolerance to defects1,2,3,4,5,6,7, organometal halide perovskites have been projected to be promising candidates for a multitude of optoelectronic applications, including photovoltaics, light emission, photodetectors, X-ray imaging, lasers, gamma-ray detection, subwavelength photonic devices in a long-wavelength region, etc.8,9,10,11,12,13,14. The rapid increase efficiency in a solar cell based on organometal halide perovskites validates its promise in photovoltaics. In the last few years, the power conversion efficiency (PCE) of mesoporous-type perovskite solar cells (PSCs) has increased to 23.3% by optimizing thin-film growth, interface, and absorber materials15,16,17. As of today, almost all PSCs with high PCE are based on mesoporous-type PSCs that often require high temperature to sinter the mesoporous layer for the best performance, compromising its low-cost advantage and limiting its application in flexible and tandem devices16,17. In order to overcome this issue, planar-type PSC comprising of stacked planar thin films has been developed18,19 using low-temperature and low-cost synthesis processes20,21,22 since the long charge diffusion length and bipolar carrier properties of perovskites23,24. However, compared to the mesoporous-type PSC, its planar-type counterpart suffers from significant lower certified PCE18,25.

In a typical planar-type PSC, the perovskite absorber usually inserts between the electron-transport layer (ETL) and the hole-transport layer (HTL) to achieve inverted p–i–n or regular n–i–p configuration21. Generally, the inverted device structure utilizing fullerene ETL displays very low hysteresis, however, it usually yields lower PCE, not to mention that fullerene is very expensive26,27. Therefore, research has focused on n–i–p architecture to provide low cost and high efficiency28,29. Even though ETL-free planar-type PSCs have been reported30,31, their highest PCE is only 14.14%, significantly lower than that of the cells with ETL, demonstrating the importance of the ETL in this configuration of PSCs. A suitable ETL should meet some basic requirements for high device efficiency32. For instance, decent optical transmittance to ensure enough light is transmitted into the perovskite absorber, matched energy level with the perovskite materials to produce the expected open-circuit voltage (Voc), and high electron mobility to extract carriers from the active layer effectively in order to avoid charge recombination, etc. Fast carrier extraction is desired to restrict charge accumulation at the interface due to ion migration for reduced hysteresis in the planar-type PSCs. Thus, emphasis has been on developing high-quality ETLs with suitable energy level and high electron mobility for high PCE devices.

Thus far, TiO2 is still the most widely used ETL in high-efficiency n–i–p planar-type PSCs due to its excellent photoelectric properties33. However, the electron mobility of TiO2 ETL is too low (ca. 10−4 cm2 V−1 s−1) to match with high hole mobility of commonly used HTLs (ca. 10−3 cm2 V−1 s−1), leading to charge accumulation at the TiO2/perovskite interface that causes hysteresis and reduced efficiency34. There have been extensive efforts in developing low-temperature TiO2 ETL, such as exploring low- temperature synthesis processes through doping and chemical engineering. The results shown by Tan et al. demonstrate that use of chlorine to modify the TiO2 microstructure at low temperatures provides promising PCE of 20.1%35. Recently, SnO2 has been demonstrated as an alternative ETL to replace TiO2, owing to its more suitable energy level relative to perovskite and higher electron mobility. Ke et al. first used SnO2 thin film as an ETL in regular planar-type PSCs and demonstrated a PCE of 16.02% with improved hysteresis36. Later, the SnO2–TiO2 (planar and mesoporous) composite layers were developed to enhance the performance of the PSCs37,38. It is noteworthy to mention that Al3+-doped SnO2 provides even better performance39. Subsequently, a variety of methods, such as solution deposition, atomic layer deposition, chemical bath deposition, etc.40,41,42 have been developed for synthesizing SnO2 thin film to improve the performance of planar-type PSCs43. Recently, Jiang et al. developed the SnO2 nanoparticles as the ETL and demonstrated a certified efficiency as high as 19.9% with very low hysteresis21. However, the PCE of the planar-type PSCs is still lower than that of the mesoporous-type devices likely due to charge accumulation at the ETL/perovskite interface caused by relatively low electron mobility of the ETL44. It is expected that better PSC performance will be achieved by increasing electron mobility of the ETLs.

Ethylene diamine tetraacetic acid (EDTA) provides excellent modification of ETLs in organic solar cells owing to its strong chelation function. Li et al. have employed EDTA to passivate ZnO-based ETL and demonstrated improved performance of the organic solar cells45. However, when the EDTA–ZnO layer is used in the present perovskite cells, the hydroxyl groups or acetate ligands on the ZnO surface react with the perovskite and proton transfer reactions occur at the perovskite/ZnO interface, leading to poor device performance46.

In the present work, we realize an EDTA-complexed SnO2 (E-SnO2) ETLs by complexing EDTA with SnO2 in planar-type PSCs to attain PCE as high as 21.60%, and certified PCE reaches to 21.52%, the highest reported value to date for the planar-type PSCs. Owing to the low-temperature processing for E-SnO2, we fabricate flexible PSCs, and the PCE reaches to 18.28%. Besides, the PSCs based on E-SnO2 show negligible hysteresis because of the eliminated charge accumulation at the perovskite/ETL interface. We find that the electron mobility of E-SnO2 increases by about three times compared to that of SnO2, leading to negligible current density–voltage (J–V) hysteresis due to improved electron extraction from the perovskite absorber21. Furthermore, we find that SnO2 surface becomes more hydrophilic upon EDTA treatment, which decreases the Gibbs free energy for heterogeneous nucleation, resulting in high quality of the perovskite film.

Results

Fabrication and characterization of E-SnO2

It is well known that EDTA can react with transition metal oxide to form a complex, because it can provide its lone-pair electrons to the vacant d-orbital of the transition metal atom47. Thus, EDTA was chosen to modify the SnO2 to improve its performance. Supplementary Fig. 1a describes the chemical reaction that occurred when the SnO2 was treated using the EDTA aqueous solution, resulting in the formation of a five-membered ring chelate. The images of EDTA, SnO2, and E-SnO2 samples are shown in Supplementary Fig. 1b. It is apparent that the unmodified EDTA and SnO2 samples are transparent, while EDTA-treated SnO2 turned into milky white. Supplementary Fig. 2 compares the Fourier-transform infrared spectroscopy (FTIR) spectra of the E-SnO2 solution measured in the freshly prepared condition and again after it was stored in an ambient atmosphere for 2 months. It is clear that there is no obvious difference between the two solutions indicating the high stability.

Figure 1a shows the X-ray photoelectron spectra (XPS) for EDTA, SnO2, and E-SnO2 films deposited on quartz substrates. In order to reduce the charging effect, the exposed surface of the quartz substrate was coated with a conductive silver paint and connected to the ground. We calibrated the binding energy scale for all XPS measurements to the carbon 1s line at 284.8 eV. It is clear from these measurements that SnO2 shows only peaks attributed to Sn and O. After the EDTA treatment, the E-SnO2 film shows an additional peak located at ca. 400 eV, ascribed to N. Meanwhile, the Sn 3d peaks from E-SnO2 are shifted by ca. 0.16 eV in contrast to the pristine SnO2 (Supplementary Fig. 3), indicating that EDTA is bound to the SnO2.

Fig. 1
figure 1

Characterization of the ETLs. a XPS and b FTIR spectra of EDTA, SnO2, and E-SnO2 films deposited on quartz substrates. c AFM topographical images of EDTA, SnO2, and E-SnO2 films. d Schematic illustration of Fermi level of EDTA, SnO2, and E-SnO2 relative to the conduction band of the perovskite layer. The Fermi level of EDTA, SnO2, and E-SnO2 is measured by KPFM, and the conduction and valence band of the perovskite materials are obtained from the previous report74. e Optical transmission spectra of EDTA, SnO2, and E-SnO2 films on ITO substrates. f Electron mobility for EDTA, SnO2, and E-SnO2 using the SCLC model, and the inset shows the device structure of ITO/Al/ETL/Al

Full size image

FTIR was used to study the interaction between SnO2 and EDTA. As shown in Fig. 1b, the peaks around 2895 cm−1 and 1673 cm−1 belong to C–H and C=O stretching vibration in the EDTA, respectively. The characteristic peaks of SnO2 observed at ca. 701 cm−1 and 549 cm−1 are due to O–Sn–O stretch and the Sn–O vibration, respectively48. In addition, the peak at 1040 cm−1 in the SnO2 film is attributed to O–O stretching vibration due to oxygen adsorption on the SnO2 surface49. For the E-SnO2 sample, the characteristic peaks of SnO2 shift to 713 cm−1 and 563 cm−1, and the C–H and C=O stretching vibration peaks shift to 2913 cm−1 and 1624 cm−1, further demonstrating that the EDTA is indeed complexed with SnO2.

Atomic force microscopy (AFM) images of EDTA, SnO2, and E-SnO2 films deposited on the ITO substrates are shown in Fig. 1c. The data reveal that the E-SnO2 film shows the smallest root-mean-square roughness of 2.88 nm, a key figure-of-merit for the PSCs50. We also measured their Fermi level by Kelvin probe force microscopy (KPFM), with the surface potential images shown in Supplementary Fig. 4, and the calculated details are described in Supplementary Note 1. Figure 1d provides energy band alignment between perovskites and different ETLs. The Fermi level of E-SnO2 is very close to the conduction band of perovskite, which is beneficial for enhancing Voc51.

Figure 1e shows the optical transmission spectra of EDTA, SnO2, and E-SnO2 films coated on ITO. All these samples display high average transmittance in the visible region, demonstrating good optical quality. In addition, the electron mobility of various ETLs was measured using the space charge-limited current (SCLC) method20, as shown in Fig. 1f. It is found that electron mobility of E-SnO2 is 2.27 × 10−3 cm2 V−1 s−1, significantly larger than those of the EDTA (3.56 × 10−5 cm2 V−1 s−1) and the SnO2 (9.92 × 10−4 cm2 V−1 s−1). It is known that the electron mobility is a key figure-of-merit for ETLs in PSCs. Supplementary Fig. 5 shows the electron injection models for ITO/SnO2 or E-SnO2/perovskite/PCBM/Al structures, with their corresponding J–V curves, and the details are described in Supplementary Note 2. It is apparent that the high electron mobility effectively promotes electron transfer in the PSCs, reduces charge accumulation at the ETL/perovskite interface, improves efficiency, and suppresses hysteresis for the PSCs21.

Perovskite growth mechanism

The quality of the perovskite films, including grain size, crystallinity, surface coverage, etc., is very important for high-performance PSCs. For a consistent microstructure, a solution deposition technique was used to fabricate perovskite films on EDTA, SnO2, and E-SnO2 substrates. Figure 2a–c shows the morphology of the perovskite films deposited on different ETLs. It is clear from these images that continuous pinhole-free films with full surface coverage were obtained. Figure 2d shows the distribution diagram with an average grain size of about 309 nm for the perovskite coated on SnO2. The grain size increased to about 518 nm for the EDTA sample. Surprisingly, the average perovskite grain size is further enhanced to as much as about 828 nm (Fig. 2c, d) for the E-SnO2 substrates.

Fig. 2
figure 2

The morphology of perovskite films deposited on different substrates. Top-view scanning electron microscope (SEM) images of perovskite films coated on a EDTA, b SnO2, and c E-SnO2 substrates. d The grain size distribution of perovskite deposited on various substrates

Full size image

According to the established model for nucleation and growth of thin films52,53, the perovskite formation process can be divided into four steps: (i) formation of a crystal nucleus, (ii) evolution of nuclei into an island structure, (iii) formation of a networked microstructure, and (iv) growth of networks into a continuous film. The Gibbs free energy for heterogeneous nucleation in the first step can be expressed as Eq. (1)

$$\bigtriangleup G_{{\mathrm{heterogeneous}}} = \bigtriangleup G_{{\mathrm{homogeneous}}} \times f\left( \theta \right)$$
(1)

wherein f(θ) = (2–3 cos θ + cos3θ)/454, and θ is the contact angle of the precursor solution. Since the magnitude of θ varies in the range of [0, π/2], the larger the θ is, the smaller is the magnitude of cos θ, and therefore larger is the parameter f(θ) ϵ [0, 1]. In other words, a smaller contact angle results in reduced Gibbs free energy for heterogeneous nucleation, thereby assisting the nucleation process. Higher nucleation density will promote the film densification process53. Compared to EDTA and SnO2, E-SnO2 shows the smallest contact angle (20.67°, Supplementary Fig. 6), resulting in the wettability interface for the perovskite55,56,57. Thus, the perovskite coated on the E-SnO2 exhibits better crystallinity (Supplementary Fig. 7) and full surface coverage (Fig. 2c). In addition, the small contact angle of the substrate provides the low surface energy58, leading to increased grain size during the growth of the networked structure53, as observed in the SEM measurements.

Charge transfer dynamics

The electron-only devices with the structure of ITO/ETL/perovskite/PCBM/Ag were fabricated to evaluate the trap density of perovskite deposited on different substrates. Figure 3a shows the dark current–voltage (I–V) curves of the electron-only devices. The linear correlation (dark yellow line) reveals an ohmic-type response at low bias voltage, when the bias voltage is above the kink point, which defines as the trap-filled limit voltage (VTFL), the current nonlinearly increases (cyan line), indicating that the traps are completely filled. The trap density (Nt) can be obtained using Eq. (2)

$$N_{\mathrm{t}} = \frac{{2\varepsilon _0\varepsilon V_{{\mathrm{TFT}}}}}{{eL^2}}$$
(2)

where ε0 is the vacuum permittivity, ε is the relative dielectric constant of FA0.95Cs0.05PbI3 (ε = 62.23)59, e is the electron charge, and L is the thickness of the film. The trap densities of the perovskite film coated on SnO2 and EDTA substrates are 1.93 × 1016 and 1.27 × 1016 cm−3, respectively. Interestingly, the trap density is reduced to as low as 8.97 × 1015 cm−3 for the film deposited on E-SnO2. The significantly lower trap density is related to low grain boundary density in the perovskite film (Fig. 2).

Fig. 3
figure 3

The charge transfer between perovskite and different ETLs. a Dark I–V curves of the electron-only devices with the VTFL kink points. The inset shows the structure of the electron-only device. b Steady-state PL and c TRPL spectra with an excitation intensity of 3 μJ cm−2 of perovskite films deposited on different substrates

Full size image

Figure 3b shows the steady-state photoluminescence (PL) spectra of the perovskite deposited on different substrates. Compared with other samples, significant PL quench is observed in the ITO/E-SnO2/perovskite, demonstrating that the E-SnO2 has the most appealing merits as the highest electron mobility (Fig. 1f). Figure 3c shows the normalized time-resolved PL (TRPL) for perovskite coated on various ETLs. The lifetime and the corresponding amplitudes are listed in Supplementary Table 1. Generally, the slow decay component (τ1) is attributed to the radiative recombination of free charge carriers due to traps in the bulk, and the fast decay component (τ2) is originated from the quenching of charge carriers at the interface60. The glass/perovskite sample shows the longest lifetime under excitation intensity of 3 μJ cm−2. For perovskite coated on the ITO substrate, the lifetime is decreased to more than half due to the charge transfer from perovskite into ITO. For EDTA/perovskite and SnO2/perovskite samples, both the fast and slow decay lifetimes are very similar, and τ1 dominates the PL decay for both samples, indicating severe recombination before they were extracted. When the perovskite is deposited on E-SnO2, both τ1 and τ2 were shortened to 14.16 ns and 0.97 ns, with a proportion of 45.32% and 54.68%, respectively. Meanwhile, τ2 appears to dominate the PL decay, indicating that electrons are effectively extracted from the perovskite layer to the E-SnO2 with minimal recombination loss. Even under smaller excitation intensity (0.5 μJ cm−2), the acceleration of the lifetime for E-SnO2/perovskite is observed. The lifetime increases with reduced excitation intensity (Supplementary Fig. 8 and Supplementary Table 1), in agreement with a previous report61. The electron-transport yield (Фtr) of different ETLs with different excitation intensities can be estimated using equation, Фtr = 1 –τp/τglass, where τp is the average lifetime for perovskite deposited on different substrates, and τglass is the average lifetime for glass/perovskite. With the excitation intensity of 3 μJ cm−2, the electron-transport yields of ITO, EDTA, SnO2, and E-SnO2 are 49.72%, 67.58%, 68.31%, and 81.50%, respectively. When the excitation intensity reduces to 0.5 μJ cm−2, the electron-transport yields of ITO, EDTA, SnO2, and E-SnO2 are increased to 60.37%, 74.46%, 80.65%, and 90.82%, respectively. It is clear that the excitation intensity can significantly increase the electron- transport yield. These results further indicate that the E-SnO2 is a good electron extraction layer for planar-type PSCs.

The performance of PSCs

With the superior optoelectronic properties discussed above, it is expected that the E-SnO2 would make a better ETL in the PSCs than the SnO2. Planar-type PSCs are therefore designed and fabricated based on different ETLs with the device structure shown in Fig. 4a inset. FAPbI3 was used as the active absorber for its proper band gap, with a small amount of Cs doping to improve its phase stability62,63. Supplementary Fig. 9 presents the cross-sectional SEM images for the complete device structure. The thickness of the perovskite film is controlled at ca. 420 nm for all devices. While the perovskite grains are not large enough to penetrate through the film thickness when the SnO2 is used as the substrate, the grains are significantly larger when deposited on EDTA and E-SnO2 with the grains grown across the film thickness, which is consistent with top-view SEM results (Fig. 2).

Fig. 4
figure 4

PSC performance using ETLs. a J–V curves with the inset showing device configuration, and b the corresponding IPCE of the planar-type PSCs with various ETLs. The integrated current density from the IPCE curves with the AM 1.5 G photon flux spectrum. c Static current density and PCE measured as a function of time for the EDTA, SnO2, and E-SnO2 devices biased at 0.85 V, 0.89 V, and 0.92 V, respectively. d The PCE distribution histogram of the planar-type PSCs based on different ETLs

Full size image

Figure 4a shows the J–V curves of planar-type PSCs using different ETLs, with the key parameters, including short-circuit current density (Jsc), Voc, fill factor (FF), and PCE summarized in Table 1. The device based on EDTA gives a PCE of 16.42% with Jsc = 22.10 mA cm−2, Voc = 1.08 V, and FF = 0.687. The device based on SnO2 substrate shows a PCE of 18.93% with Jsc = 22.79 mA cm−2, Voc = 1.10 V, and FF = 0.755. Interestingly, when the E-SnO2 is employed as ETL, the Jsc, FF, and Voc are increased to 24.55 mA cm−2, 0.792, and 1.11 V, yielding a PCE up to 21.60%, (the certified efficiency is 21.52%, and the certificated document is shown in Supplementary Fig. 10), the highest efficiency reported to date for the planar-type PSCs. The low device performance for the EDTA is caused by small Jsc and FF, which is related to low electron mobility and high resistance47, and the low Voc results from the small offset of Fermi energy between the EDTA and HTL (Fig. 1d)64. In comparison, the planar-type PSCs with E-SnO2 ETLs exhibit the best performance. The higher Jsc and FF are attributed to the high electron mobility that promotes effective electron extraction, and the larger Voc due to the closer energy level between E-SnO2 and perovskite65. Figure 4b shows the incident-photon-to-charge conversion efficiency (IPCE) and the integrated current of the PSCs based on different ETLs. The integrated current values calculated by the IPCE spectra for the devices using EDTA, SnO2, and E-SnO2 are 21.22, 21.58, and 24.15 mA cm−2, respectively, very close to the J–V results. It is apparent that the device based on the E-SnO2 shows significantly higher IPCE due to less optical loss when perovskite is deposited on E-SnO2 ETL (Supplementary Fig. 11), consistent with the J–V measurement.

Table 1 The parameters of the rigid and flexible devices
Full size table

To further demonstrate the device characteristics, photocurrent density of the champion devices from each group based on EDTA, SnO2, and E-SnO2 was measured when the devices were biased at 0.85, 0.89, and 0.92 V, respectively. Figure 4c shows the corresponding curves at the maximum power point (Vmp) in the J–V plots. The PCEs of the champion devices using the EDTA, SnO2, and E-SnO2 stabilize at 16.34%, 18.67%, and 21.67% with photocurrent densities of 19.22, 20.98, and 23.55 mA cm−2, respectively, very close to the values measured from the J–V curves. Next, we fabricated and measured 30 individual devices for each ETL to study repeatability. Figure 4d shows the PCE distribution histogram for devices with different ETLs, with the statistics listed in Supplementary Tables 2–4. Amazingly, the devices based on E-SnO2 exhibit excellent repeatability with a very small standard deviation in contrast to the devices based on EDTA and SnO2, indicating that the E-SnO2 is an excellent ETL in the planar-type PSC.

In order to gain further insight into the charge transport mechanism, the charge transfer processes in the perovskite devices were studied in detail. The carrier recombination rate in the PSCs was evaluated by the Voc decay measurements. Figure 5a shows the Voc decay curves of the PSCs based on different ETLs. It is apparent that the planar-type PSC based on E-SnO2 exhibits the slowest Voc decay time compared to the devices based on EDTA and SnO2, indicating that the devices with E-SnO2 have the lowest charge recombination rate and the longest carrier lifetime, consistent with the highest Voc for the device based on E-SnO2 by J–V measurements. Figure 5b shows Jsc versus light intensity of the PSCs using various ETLs. It appears that all devices show a linear correlation with the slopes very close to 1, indicating that the bimolecular recombination in the devices is negligible66. Figure 5c shows that Voc changes linearly with the light intensity. Prior studies have indicated that the deviation between the slope and the value of (kT/q) reflects the trap-assisted recombination20. In the present case, the device using the E-SnO2 shows the smallest slope, indicating the least trap-assisted recombination, which is in excellent agreement with the result showing the lowest trap density when the perovskite is deposited on E-SnO2 (Fig. 3a). In fact, the slope is as small as 1.02 kT/q, implying that the trap-assisted recombination is almost negligible.

Fig. 5
figure 5

Charge transfer properties of the planar-type PSCs using different ETLs. a Voc decay curves, b Jsc vs. light intensity, c Voc vs. light intensity, and d EIS of planar-type PSCs with various ETLs

Full size image

The electrical impedance spectroscopy (EIS) was employed to extract transfer resistance in the solar cells. Figure 5d shows the Nyquist plots of the devices using different ETLs measured at Voc under dark conditions, with the equivalent circuit shown in Supplementary Fig. 12. It is known that in the EIS analysis, the high-frequency component is the signature of the transfer resistance (Rtr) and the low-frequency one for the recombination resistance (Rrec)67. In the present study, because the perovskite/HTL interface is identical for all devices, the only variable affecting Rtr is the perovskite/ETL interface. The numerical fitting gives the device parameters, as listed in Supplementary Table 5. Apparently, compared to PSCs based on EDTA and SnO2, the device with E-SnO2 shows the smallest Rtr of 14.8 Ω and the largest Rrec of 443.3 Ω. The small Rtr is beneficial for electron extraction, and the large Rrec effectively resists charge recombination, which is in agreement with the observations discussed above. Combined, all the results confirm that E-SnO2 is the most effective ETL for the planar-type PSC.

Stability and hysteresis

Stability and hysteresis are two key characteristics for the PSCs. Figure 6a shows normalized PCE measured as a function of storage time, with more detailed J–V parameters summarized in Supplementary Table 6. It is clear that while the device based on E-SnO2 maintains 92% of its initial efficiency exposed to an ambient atmosphere after 2880 h in the dark, the device using SnO2 only provides 74% of its initial efficiency under the same storage condition. The PSCs were also tested under continuous irradiation at 100 mW cm−2. Figure 6b shows the normalized PCE changes as a function of test time, with more detailed J–V parameters provided in Supplementary Table 7. It is clear that after 120 h of illumination, the device using the E-SnO2 maintains 86% of its initial efficiency, while for the same test duration, the device using SnO2 remains only 38% relative to its initial efficiency. It is apparent that the device fabricated on E-SnO2 shows excellent stability under both the dark and continuous irradiation. The instability of PSC is mainly caused by degradation of the perovskite film and spiro-OMeTAD HTL. In the present work, all devices used the same spiro-OMeTAD HTL, therefore, the degradation from the HTL should be the same for all the devices. It is found that the grain size of the perovskite film is increased by three times when it is deposited on E-SnO2 in comparison to that on the pristine SnO2 (Fig. 2). The larger grain size can effectively suppress the moisture permeation at grain boundaries68, resulting in improved environmental stability for the PSCs based on the E-SnO2 ETLs.

Fig. 6
figure 6

Stability and hysteresis test for planar-type PSCs. Long-term stability measurements of devices without any encapsulation under a ambient condition and b illumination of 100 mW cm−2. The J–V curves of the device with c SnO2 and d E-SnO2 measured under both reverse- and forward-scan directions

Full size image

For the hysteresis test, Fig. 6c and d show the J–V curves measured under both reverse- and forward- scan directions. It is found that the device with E-SnO2 has almost identical J–V curves with negligible hysteresis, even when it is measured using different scan rates from 0.01 to 0.5 V s−1. Supplementary Fig. 13 presents J–V curves measured for the device based on E-SnO2 at different scan rates. It is apparent that the J–V curves almost remain the same, regardless of scan rate and direction, demonstrating that the hysteresis is negligible. Generally, the hysteresis of PSCs is ascribed to interfacial capacitance caused by charge accumulation at the interface, which originates from ion migration, high trap density, and unbalanced charge transport within the perovskite device69,70,71. It is found that the trap density of the perovskite film is significantly reduced when it is deposited on the E-SnO2, one of the primary reasons for reduced hysteresis. In addition, the electron mobility of the SnO2 ETL is only 9.92 × 10−4 cm2 V−1 s−1 (Fig. 1f), about an order of magnitude slower than the hole mobility of the doped spiro-OMeTAD (ca. 10−3 cm2 V−1 s−1) HTL. Thus, the electron flux (Fe) is ca. 10 times smaller than the hole flux (Fh) due to the same interface area of the ETL/perovskite and perovskite/HTL, that leads to charge accumulation at the SnO2/perovskite interface, as shown in Supplementary Fig. 14a. The accumulated charge would cause hysteresis in the solar cells (Fig. 6c). When the high electron mobility E-SnO2 (2.27 × 10−3 cm2 V−1 s−1) is employed as the ETL, the Fe is comparable to the Fh of the spiro-OMeTAD HTL (Supplementary Fig. 14b), resulting in equivalent charge transport at both electrodes. Therefore, the high electron mobility of E-SnO2 would enhance electron transport from perovskite to E-SnO2 ETL, leading to no significant charge accumulation, and consequently, the devices based on the E-SnO2 exhibit negligible hysteresis.

High-efficiency flexible PSCs

Given the advantage of low-temperature preparation, we applied the E-SnO2 ETL in flexible PSCs. Figure 7a shows J–V curves of flexible PSCs using the poly(ethylene terephthalate) (PET)/ITO substrates, with key J–V parameters summarized in Table 1. The champion flexible device exhibits PCE of 18.28% (Jsc = 23.42 mA cm−2, Voc = 1.09 V, and FF = 0.716). The lower Jsc of the flexible device is caused by the lower transparency of the PET/ITO substrate compared to the glass/ITO used for the rigid device (Supplementary Fig. 15). The lower Voc and FF are likely due to higher sheet resistance of the PET/ITO substrate67. Figure 7c shows the IPCE and integral current density of the flexible device. It is clear that the integral current is 23.12 mA cm−2, in perfect agreement with the J–V results. For the reproducibility test, 30 individual cells were fabricated with the PCE distribution histogram shown in Fig. 7d and detailed parameters are summarized in Supplementary Table 8, both confirming very good reproducibility.

Fig. 7
figure 7

The performance of flexible PSCs based on E-SnO2 ETLs. a J–V curves of the flexible devices and after flexing at curvature radii of 14 mm, 12 mm, and 7 mm for 500 cycles, respectively. b The normalized PCE measured after flexing at different curvature radii. c IPCE curves of the flexible device. d The PCE distribution histogram of the flexible PSCs

Full size image

The mechanical stability is an important quality indicator for the flexible solar cells. According to a previous report72, it is safe for ITO to be bended to a radius of 14 mm, and when the bending radius is smaller than 14 mm, the ITO layer starts to crack, leading to significant degradation in conductivity. In order to examine the mechanical stability of the flexible PSCs, we therefore adopted the bending radii of 14 mm, 12 mm, and 7 mm to test the flexible device. Figure 7a shows device performance of the flexible solar cells measured after flexing for 500 times with different curvature radii, and the test procedure is shown in Fig. 7b. It shows that after flexing for 500 times at a bending radius of 14 mm, the J–V curve and the associated parameters remain the same without obvious degradation. However, when the bending radius is decreased to 12 mm and 7 mm, the PCE degraded to 17.82% and 16.84%, respectively, attributing to the conductivity degradation of ITO72.

Discussion

An effective E-SnO2 ETL has been developed, and the PCE of planar-type PSCs is increased to 21.60% with negligible hysteresis, and the certified efficiency is 21.52%, this is the highest reported value for planar-type PSCs so far. By taking advantage of low-temperature processing for E-SnO2 ETLs, flexible devices with high PCE of 18.28% are also fabricated. The significant performance of the planar-type PSCs is attributed to the superior advantages when perovskite is deposited on E-SnO2 ETLs, including larger grain size, lower trap density, and good crystallinity. The higher electron mobility facilitates electron transfer for suppressed charge accumulation at the interface, leading to high efficiency with negligible J–V hysteresis. Furthermore, the long-term stability is significantly enhanced since the large grain size that suppressed perovskite degradation at grain boundaries. This work provides a promising direction toward developing high-quality ETLs, and we believe that the present work will facilitate transition of perovskite photovoltaics.

Methods

Materials

NH2CHNH2I (FAI) was synthesized and purified according to a reported procedure45. The SnO2 solution was purchased from Alfa Aesar (tin (IV) oxide, 15 wt% in H2O colloidal dispersion). PbI2 (purity > 99.9985%) was purchased from Alfa Aesar. EDTA (purity > 99.995%), CsI (purity > 99.999%), dimethylformamide (DMF, purity > 99%), and dimethyl sulfoxide (DMSO, purity > 99%) were obtained from Sigma Aldrich. In total, 2,2′,7,7′-tetrakis(N,N-di-p-methoxyphenylamine)-9,9′-spirobifluorene (spiro-OMeTAD) was bought from Yingkou OPV Tech Co., Ltd. All of the other solvents were purchased from Sigma Aldrich without any purification.

Fabrication of EDTA, SnO2, and E-SnO2 films

The 0.2-mg EDTA was dissolved in 1 mL of deionized water, and the SnO2 aqueous colloidal dispersion (15 wt%) was diluted using deionized water to the concentration of 2.5 wt%. These solutions were stirred at room temperature for 2 h. The SnO2 and EDTA layers were fabricated by spin-coating at 5000 rpm for 60 s using the corresponding solution, and then dried in a vacuum oven at 60 °C under ca. 5 Pa for 30 min to remove residual solvent. The EDTA and SnO2 solution were mixed with a volume ratio of 1:1, then put on a hot plate at 80 °C for 5 h under stirring conditions, and the milky-white E-SnO2 colloidal solution (Supplementary Fig. 1b) was obtained. The E-SnO2 colloidal solution was spin-coated at 5000 rpm for 60 s, and then transferred the samples into a vacuum oven at 60 °C under ca. 5 Pa for 30 min to remove the residual solvent. Finally, the E-SnO2 films were obtained.

Electron mobility of EDTA, SnO2, and E-SnO2 films

To gain insights into the charge transport, we have measured electron mobility using different ETLs in the same device structure. Specifically, the electron-only device was designed and fabricated using ITO/Al/ETL/Al structure, as shown in the inset in Fig. 1f. In this analysis, we assumed that the current is only related to electrons. When the effects of diffusion and the electric field are neglected, the current density can be determined by the SCLC73. The thickness of 80-nm Al was deposited on ITO substrates by thermal evaporation, and then the different ETLs were spin-coated on ITO/Al. Finally, 80-nm-thick Al was deposited on ITO/Al/ETL samples. The dark J–V curves of the devices were performed on a Keithley 2400 source at ambient conditions. The electron mobility (μe) is extracted by fitting the J–V curves using the Mott–Gurney law (3)

$$\mu _{\mathrm{e}} = \frac{{8JL^3}}{{9\varepsilon _0\varepsilon \left( {V_{{\mathrm{app}}} - V_{\mathrm{r}} - V_{{\mathrm{bi}}}} \right)^2}}$$
(3)

where J is the current density, L the thickness of different ETLs, ε0 the vacuum permittivity, εr the dielectric permittivity of various ETLs, Vapp the applied voltage, Vr the voltage loss due to radiative recombination, and Vbi the built-in voltage owing to the different work function between the anode and cathode.

Fabrication of solar cells

The perovskite absorbers were deposited on different ETL substrates using one-step solution processed. In total, 240.8 mg of FAI, 646.8 mg of PbI2, and 18.2 mg of CsI were dissolved in 1 mL of DMF and DMSO (4:1, volume/volume), with stirring at 60 °C for 2 h. The precursor solution was spin-coated on the EDTA, SnO2 and E-SnO2 substrates. The spin-coated process was divided by a consecutive two-step process, the spin rate of the first step is 1000 rpm for 15 s with accelerated speed of 200 rpm, and the spin rate of the second step is 4000 rpm for 45 s with accelerated speed of 1000 rpm. During the second step end of 15 s, 200 μL of chlorobenzene was drop-coated to treat the perovskite films, and then the perovskite films were annealed at 100 °C for 30 min in a glovebox. After cooling down to room temperature, the spiro-OMeTAD solution was coated on perovskite films at 5000 rpm for 30 s with accelerated speed of 3000 rpm. The 1-mL HTL chlorobenzene solution contains 90 mg of spiro-OMeTAD, 36 μL of 4-tert-butylpyridine, and 22 μL of lithium bis(trifluoromethylsulfonyl) imide of 520 mg mL−1 in acetonitrile. The samples were retained in a desiccator overnight to oxidate the spiro-OMeTAD. Finally, 100-nm-thick Au was deposited using thermal evaporation. The device area of 0.1134 cm2 was determined by a metal mask.

Characterization

The J–V curves of the PSCs were measured using a Keithley 2400 source under ambient conditions at room temperature. The light source was a 450-W xenon lamp (Oriel solar simulator) with a Schott K113 Tempax sunlight filter (Praezisions Glas & Optik GmbH) to match AM1.5 G. The light intensity was 100 mW cm−2 calibrated by a NREL-traceable KG5-filtered silicon reference cell. The active area of 0.1017 cm2 was defined by a black metal aperture to avoid light scattering into the device, and the aperture area was determined by the MICRO VUE sol 161 instrument. The J–V curves for PSCs were tested both at reverse scan (from 2 to −0.1 V, step 0.02 V) and forward scan (from −0.1 to 2 V, step 0.02 V), and the scan rate was selected from 0.01 to 0.5 V s−1. There was no preconditioning before the test. The IPCE was implemented on the QTest Station 2000ADI system (Crowntech. Inc., USA). AFM height images were attained by a Bruker Multimode 8 in tapping mode. KPFM was carried out on Bruker Metrology Nanoscope VIII AFM in an ambient atmosphere. The TRPL spectra were performed on an Edinburgh Instruments FLS920 fluorescence spectrometer. SEM images were gained by a field-emission scanning electron microscope (SU8020) under an accelerating voltage of 2 kV. XPS measurements were performed on an AXISULTRA X-ray photoelectron spectrometer. The optical transmission was acquired by a Hitachi U-3900 spectrophotometer.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.