Introduction

Organic–inorganic lead halide perovskites have many excellent advantages, such as high absorption coefficient, bandgap tunability, long carrier diffusion length, high hole and electron mobilities, and compatibility with low-cost solution processing1,2,3,4,5,6. By taking these advantages, perovskites are used as the light absorber of solar cells. The highest certified power conversion efficiency (PCE) of perovskite solar cells (PSCs) has already reached 25.2%7, which is comparable to silicon solar cell technology. However, because perovskites are composed of many kinds of ions, the ion migration through PSCs is a potential obstacle for efficient and stable solar power conversion in PSCs. The ion migration frequently causes current density–voltage (JV) hysteresis and reduced operational stability8,9,10,11. Various approaches have been employed to reduce the ion migration. The reported approaches include compositional engineering of perovskite materials12, reducing grain boundary density by increasing grain size13,14, and passivation of grain boundaries using organic materials6,13. However, the current situation is still far from completely blocking the ion migration in perovskite materials and devices11,15. It is also necessary to clarify mechanisms of the ion migration behind.

Weber et al.16 recently reported that the localization of positively charged ions at the interface between the perovskite layer and the electron transport layer (ETL) causes the JV hysteresis in PSCs16. The localized positive ions accelerate non-radiative recombination, leading to the quick degradation of PSCs as well16,17,18. Here, the positive ions are defined as all possible ions and vacancies. However, why the localization of the positive ions happens is still unclear.

Herein, we report that the presence of hydroxyl (–OH) groups at a metal oxide ETL is one possible origin of this ionic localization. Metal oxides like SnO2 used in this study as the ETL typically have the –OH groups on the surfaces19. To avoid the localization of the positive ions, we deactivate the –OH groups by chemically modifying the SnO2 surface with a self-assembled monolayer (SAM) of a fullerene derivative. Introducing this interfacial SAM significantly reduces the JV hysteresis. With this SAM treatment, almost no degradation of PSCs is obtained at room temperature under continuous illumination for 1000 h.

Results and discussion

Ion migration and hysteresis

To identify the existence of the ion migration, temperature-dependent electrical conductivity measurements were carried out by measuring current–voltage (IV) curves of perovskite films in a lateral-electrode structure. This measurement was designed to highlight ionic conduction, as ions would not contribute to the conductivity at low temperatures, but would contribute when the temperature was sufficiently high to provide the energy required to form mobile ions20. Figure 1a shows IV curves and revealed that the hysteresis increased with increasing temperature. This result indicates the existence of the ion migration in our perovskite films and is consistent with a previous report21,22,23,24.

Fig. 1: Indication of ion migration in perovskite films and perovskite solar cells.
figure 1

a Temperature-dependent conductivity measurements in a lateral structured perovskite film. b Scan-rate-dependent current density–voltage curves of perovskite solar cells. c, d Effect of temperature on the hysteresis of perovskite solar cells at a fast scan rate of 200 mV s‒1. Current density–voltage curves measured at 60 °C (c) and after cooling to room temperature (d).

We fabricated planar PSCs with a regular structure of glass substrate/indium tin oxide (ITO)/SnO2/perovskite/2,2′,7,7′-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifluorene (spiro-OMeTAD)/Au and measured their JV curves. Here we defined the voltage scan from −0.1 to 1.2 V as the forward scan and the voltage scan from 1.2 to −0.1 V as the reverse scan. In this study, we carried out the consecutive forward-to-reverse scans for the PSC evaluation, unless otherwise mentioned. Eight devices fabricated in the same batch had the average PCEs of 17.4% and 17.7% in the forward and reverse scans, respectively, at the scan rate of 200 mV s−1 (Supplementary Table 1). The JV hysteresis seemed very small at this scan rate, and the hysteresis index (HI) defined in the “Methods” section was close to zero. This small hysteresis is consistent with literature, in which SnO2 was used as the ETL in PSCs25,26. However, the hysteresis increased with decreasing the scan rates from 200 to 10 mV s−1 (Fig. 1b and Supplementary Table 2). The negative HI values obtained at the lower scan rates mean “inverted hysteresis,” for which the PCE is higher in the forward scan than in the reverse scan.

It has been reported that one of the relevant origins of the inverted hysteresis is the ion migration in PSCs9,27,28. The migration of different ionic species in perovskite films could cause the JV hysteresis, the degree of which depends on the voltage scan rates9, directions (forward-to-reverse or reverse-to-forward bias scans)29, ranges27,30, and temperature31. Thus, we measured JV curves in the consecutive reverse-to-forward scans (Supplementary Fig. 1a), which are opposite to the forward-to-reverse scans we used earlier (Supplementary Fig. 1b). The scan rate was 10 mV s−1. In this case, the reverse scan is better than the forward scan in terms of the PCE. Eventually, the scan directions are not so important; the first scan provides the higher PCE and repeating the scans decreased the PCEs. Next, we measured JV curves in a wider scan range between −0.1 and 2.0 V at a scan rate of 200 mV s−1 (Supplementary Fig. 2). The hysteresis began to appear with this scan range, although the hysteresis was not clearly observed when we used the small scan range from −0.1 to 1.2 V (Fig. 1b). This may mean that applying a high voltage to PSCs caused more serious hysteresis. For a better understanding of the temperature effect on the hysteresis, we measured JV curves at a high temperature of 60 °C at a scan rate of 200 mV s−1, resulting in the larger JV hysteresis (Fig. 1c). The hysteresis again decreased after cooling the PSCs to room temperature (Fig. 1d). Considering the aforementioned experimental results, the ion migration is likely in our PSCs.

Detection of –OH groups on a SnO2 surface

We speculate that the –OH groups attract the positive ions existing near the SnO2 surface in our PSCs, although most of the positive ions move to the opposite spiro-OMeTAD side along the internal built-in field. The –OH groups were previously identified from the SnO2 surface32,33. To observe the –OH groups, we characterized the SnO2 surface with X-ray photoelectron spectroscopy (XPS). The Sn 3d core level peak was detected (Fig. 2a). By analyzing the O 1s core level region (Fig. 2b), an XPS peak originating from the –OH groups was identified at 531.7 eV near a main oxygen peak at 530.2 eV. This is proof of the presence of the –OH groups on the SnO2 surface19,34,35, which is expected to induce the localization of the positive ions.

Fig. 2: Presence of surface charges on a SnO2 surface induces ion accumulation and, therefore, increases carrier recombination in perovskite solar cells.
figure 2

a, b X-ray photoelectron spectroscopy peaks of a SnO2 thin film in the Sn 3d (a) and O 1s (b) regions. c Electrochemical impedance spectroscopy results of perovskite solar cells. The electrochemical impedance spectroscopy measurements were repeated three times. d Illustration showing the positive ion localization caused by the surface-negative charges (–OH) of SnO2. The possible mobile positive ions are MA+ (methylammonium ion), FA+ (formamidinium ion), VI+ (iodine vacancy), and VBr+ (bromine vacancy).

Electrochemical impedance spectroscopy (EIS) measurements were carried out on the PSCs to characterize the interfacial charge dynamics caused by the localized positive ions. Figure 2c shows EIS data recorded at an open-circuit voltage (VOC) condition under illumination. The EIS measurements consisted of a series of three consecutive measurements to check the consistency of the results during the entire EIS measurement procedure. The high-frequency semicircle for every consecutive spectrum increased. The low-frequency impedance did not have a clear pattern and did not change in shape or magnitude throughout the EIS measurement procedure. The increase in impedance at 0.07 Hz (lowest measured frequency) from 12.2 to 13.8 Ω cm2 was predominantly caused by the high-frequency semicircle. This feature was previously attributed to poor electron extraction or recombination of charge carriers at the interfaces between the charge transport layer and the perovskite layer36. Recently, we reported that the short-term increase in high-frequency impedance is ~100% reversible by keeping the PSC in the dark37. This feature was related to the ion migration and accumulation, which leads to the increased non-radiative recombination37. Therefore, we attributed the short-term increase in high-frequency impedance in Fig. 2c to the localized ionic charges at the negatively charged surface of SnO2. The localized and accumulated ions, including charged vacancies (especially, iodine vacancies), are well known to act as trap-assisted charge recombination centers in PSCs17,28,38,39. Therefore, electron extraction to SnO2 was reduced because of the trap-assisted recombination, as depicted in Fig. 2d, resulting in the large hysteresis.

Surface treatment with a SAM

To deactivate the surface –OH groups, the SnO2 layer was modified by a SAM of a fullerene (C60) derivative (Fig. 3a). The C60 derivative used here was C60 pyrrolidine tris-acid (CPTA) because this material has the ability to form strong chemical bonding with the SnO2 surface40,41 and to behave as a good electron acceptor. The CPTA-SAM was formed on the SnO2 surface by spin coating from N,N-dimethylformamide (DMF) solution. The existence of the CPTA-SAM was confirmed by the appearance of XPS peaks from the C 1s and N 1s core levels and the decreased intensities of XPS peaks from the Sn 3d and O 1s core levels (Supplementary Fig. 3). Even after washing the substrates with DMF, the intensities of all the XPS peaks from the C 1s, N 1s, Sn 3d, and O 1s core levels were unchanged (Supplementary Fig. 3), indicating that most of the CPTA molecules are chemically bounded on the SnO2 surface. With the CPTA-SAM formation, the –COO– groups were formed through the esterification reaction40,42 between the CPTA molecules and the surface –OH groups. This was supported by the appearance of an XPS peak originating from the –COO– groups (Supplementary Fig. 4). In addition, the amount of the –OH groups at the SnO2 surface was reduced by ~60% after the CPTA-SAM formation (Supplementary Fig. 4).

Fig. 3: Positive effect of using a self-assembled monolayer (SAM) on the hysteresis and carrier recombination characteristics in perovskite solar cells.
figure 3

Illustration showing a a self-assembled monolayer formed on the SnO2 surface and b reduced localized ionic charges at the surface of SnO2 because of the deactivation of surface –OH groups. c Scan-rate-dependent current density–voltage curves of perovskite solar cells with a self-assembled monolayer at room temperature. d Current density–voltage curves measured at a fast scan rate of 200 mV s−1 at 60 °C. e Electrochemical impedance spectroscopy results of perovskite solar cells with a self-assembled monolayer. f Scan-rate-dependent current density–voltage curves of perovskite solar cells with a C60 layer at room temperature.

To check the influence of deactivating the –OH groups by the CPTA-SAM on the JV hysteresis, we fabricated PSCs in the presence and absence of the SAM treatment. PSCs without the CPTA-SAM had the pronounced JV hysteresis (Supplementary Fig. 5). This was consistent with the previous result shown in Fig. 1b. In contrast, adding the CPTA-SAM significantly reduced the JV hysteresis of PSCs (Fig. 3b and Supplementary Table 3). The average PCEs of eight devices with the CPTA-SAM were 18.06% and 18.12% in the forward and reverse scans, respectively (Supplementary Table 4). In addition, no hysteresis was observed in PSCs with the CPTA-SAM even at a high temperature of 60 °C (Fig. 3c). Furthermore, there was no increase in the size of the semicircle in EIS results shown in Fig. 3d.

The improved electron extraction to SnO2 by the use of the CPTA-SAM is likely for the reduced hysteresis on the basis of literature43,44. Therefore, we prepared additional PSCs using a thin layer of pure C60. Using this C60 layer can distinguish whether the decreased hysteresis resulted from the improved electron extraction or reduced localization of the positive ions because the surface –OH groups still exist below a C60 layer. As shown in Supplementary Fig. 6, the PL intensity reduction was similar when a perovskite film was prepared on the C60 layer or the CPTA-SAM, indicating the similar electron extraction efficiency in both the cases. The reduced PL intensities also pointed to the existence of C60 even after the spin coating of the perovskite film. The pronounced JV hysteresis was observed when C60 was used in PSCs (Fig. 3e and Supplementary Tables 5 and 6), despite the similar electron extraction. This indicates that the deactivation of the surface –OH groups is very important for reducing the hysteresis. Furthermore, the CPTA-SAM and the C60 layer could have different dipole moments, which may affect the ion localization and the PSC performance.

In addition to the ion migration, unbalanced charge transport45,46,47, ferroelectricity47,48, capacitive effect47,49, and defect-induced electronic traps50 are known to induce the JV hysteresis. There is a possibility that all these factors changed in our PSCs simultaneously when the SnO2 surfaces were treated differently, although every perovskite film fabricated on the untreated, CPTA-treated, and C60-coated surfaces had similar morphologies (Supplementary Fig. 7). We will carry out experiments in our laboratory to clarify the more detailed origins of the JV hysteresis.

Long-term stability under continuous light illumination

We measured the stability of PSCs with and without the CPTA-SAM under continuous illumination with an intensity of 100 mW cm−2 near a maximum power point. Figure 4a shows the PCE evolution of PSCs over time, and other parameters of JSC, VOC, and fill factor (FF) are shown in Supplementary Fig. 8. The PCEs of PSCs without the CPTA-SAM significantly dropped. It is worth mentioning here that PSCs with the CPTA-SAM had no decrease of the PCE after 1000 h of the continuous illumination. We also investigated the stability of PSCs with the C60 layer. There was no improvement in stability when using C60 (Supplementary Fig. 9). Furthermore, we obtained the improved stability with the CPTA-SAM under the illumination even at a high temperature of 60 °C (Fig. 4b and Supplementary Fig. 10). These results clearly demonstrated the importance of using the surface treatment for fabricating stable PSCs.

Fig. 4: Addition of a self-assembled monolayer (SAM) improves long-term stability of perovskite solar cells.
figure 4

Efficiency evolution of perovskite solar cells at room temperature (25 °C) (a) and at elevated temperature (60 °C) (b).

In summary, we found that the JV hysteresis and degradation can be suppressed by chemically modifying the SnO2 surface with a SAM. We speculated that one reason for the suppressed hysteresis and degradation is the deactivation of the surface –OH groups, which attract the positive ions. We obtained almost no degradation in PSCs with CPTA-SAM after 1000 h of continuous illumination at room temperature. Additionally, the introduction of CPTA-SAM greatly improved the high-temperature stability. These findings will provide insights into mechanisms of the hysteresis and long-term stability of PSCs.

Methods

Materials

A SnO2 colloidal precursor solution [tin(IV) oxide, 15% in H2O colloidal dispersion] was purchased from Alfa Aesar. CPTA was purchased from Sigma-Aldrich. All precursor materials, such as cesium iodide (CsI), formamidinium iodide (FAI), methylammonium bromide (MABr), lead iodide (PbI2), and lead bromide (PbBr2), for the perovskite fabrication were purchased from Tokyo Chemical Industry. Spiro-OMeTAD was purchased from Merck. The solvents, such as chlorobenzene, DMF, and dimethyl sulfoxide (DMSO), and the dopants for the hole transport material, such as lithium bis(trifluoromethanesulfonyl)imide (LiTFSI), 4-tert-butylpyridine (4-tBP), and FK-209 (tris(2-(1H-pyrazol-1-yl)-4-tert-butylpyridine)cobalt(III) tri[bis(trifluoromethane)sulfonimide]), were purchased from Sigma-Aldrich. These purchased materials were used as received without further purification.

Device fabrication

Glass substrates coated with a pre-patterned ITO layer with a thickness of 100 nm (Atsugi Micro, Japan) and a sheet resistance of 10 Ω sq−1 were cleaned sequentially by ultrasonication in detergent, pure water, acetone, and isopropanol for 10 min each and then subjected to ultraviolet (UV)–ozone treatment for 15 min. Next, a SnO2 film was deposited on the ITO surface by spin coating at 3000 r.p.m. for 30 s using aqueous SnO2 colloidal solution, which was diluted to 2.67% in water. This SnO2 film was annealed at 150 °C for 30 min to dry and treated with UV–ozone for 15 min. Then, the CPTA-SAM was formed on the SnO2 surface by spin coating at 4000 r.p.m. for 30 s from CPTA solution in DMF with a concentration of 0.5 mg ml−1. To remove physiosorbed CPTA molecules from the SnO2 surface, the substrate was washed by DMF and dried at 100 °C for 5 min. The CsFAMA perovskite layer was prepared according to the literature4. Specifically, precursor solution used here was prepared by mixing FAI (1.10 M), PbI2 (1.16 M), MABr (0.2 M), and PbBr2 (0.2 M), each of which was dissolved in anhydrous DMF/DMSO (4:1 vol/vol). Additionally, 0.08 M CsI solution in DMSO was added into the aforementioned precursor solution. The precursor solution was stirred at 70 °C for 3 h and then filtered with a 0.2 µm polytetrafluoroethylene filter before use. Using this solution, a perovskite film was spin coated on the untreated and treated SnO2 surfaces at 1000 r.p.m. for 10 s and then at 6000 r.p.m. for 30 s. Ten seconds before the end of the substrate rotation, 120 µl of chlorobenzene was dropped onto the spinning substrate. Subsequently, the perovskite layer was annealed at 100 °C for 45 min. The spiro-OMeTAD layer was deposited on top of the perovskite layer by spin coating at 4000 r.p.m. for 30 s from solution of spiro-OMeTAD (36.5 mg in 0.5 ml chlorobenzene). This solution was doped with 15 µl of 4-tBP stock solution (520 mg ml–1 in acetonitrile), 9 µl of LiTFSI stock solution (520 mg ml–1 in acetonitrile), and 14.5 µl of FK-209 stock solution (300 mg ml–1 in acetonitrile). The preparation of perovskite and spiro-OMeTAD was done in a nitrogen-filled glove box, while the SnO2 films were fabricated in air. Finally, a 100-nm-thick gold electrode was thermally deposited in vacuum to complete the devices. The deposition rate was 0.1 nm s−1 and the base pressure was 10−4 Pa for the Au deposition. The fabricated PSCs were encapsulated using a glass lid and UV-cured sealant.

Characterization

JV curve measurements were performed on the PSCs using a computer-controlled Keithley 2400 source unit under simulated AM1.5 G solar illumination from a Xe lamp-based solar simulator (SRO-25 GD, Bunko-Keiki, Japan) with a scan rate of 200, 100, 20, or 10 mV  s–1. EIS measurements were performed using a Solartron 1260 frequency–response analyzer from 1 MHz to 70 mHz at an excitation amplitude of 20 mV. While the original PSC area defined by the overlap of the ITO and Au electrodes was 4 mm2, the illumination area during the JV measurements was defined at 3.24 mm2 (1.8 × 1.8 mm2) using a black shadow mask. The lamp power was carefully calibrated at 100 mW cm−2 (1 Sun) using a crystalline Si reference cell with an amorphous Si optical filter (Bunko-Keiki), which was certificated by the National Institute of Advanced Industrial Science and Technology of Japan.

The HI was calculated by the following formula:

$${\mathrm{HI}} = \frac{{{\mathrm{PCE}}_{{\mathrm{reverse}}} - {\mathrm{PCE}}_{{\mathrm{forward}}}}}{{{\mathrm{PCE}}_{{\mathrm{reverse}}}}},$$

where PCEreverse and PCEforward are the PCE values measured in the forward and reverse bias scans, respectively. The calculated HI values are summarized in Supplementary Tables 110.

Device lifetime measurements

Light from a white light-emitting diode array with an intensity of 100 mW cm−2 was continuously illuminated on the encapsulated PSCs in ambient air at room temperature (~25 °C) or 60 °C. The evolution of VOC, JSC, FF, and PCE was automatically measured using a lifetime measurement system (System Engineers). In between scans, the PSC was connected with a 1 kΩ resistor22.

Steady-state PL measurements

PL spectra of perovskite films were obtained from a JASCO FP-8600 spectrofluorometer in the wavelength range from 600 to 900 nm with a wavelength step of 0.2 nm. The excitation wavelength used here was 460 nm.

Temperature-dependent conductivity measurements

Temperature-dependent conductivity measurements were carried out using a probe station (Lakeshore, TTP4) equipped with an environmental chamber and thermoelectric plate. The temperature of the device was controlled by either heating with a thermoelectric plate or cooling with a flow of liquid nitrogen. Electrical measurements were conducted using a source/measurement unit (Agilent, B2902A). A lateral device with a configuration of Au/perovskite (100 μm)/Au was used for the measurement. The distance between the two Au electrodes was 100 μm.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.