Recently, a record power conversion efficiency (PCE) of perovskite/silicon monolithic tandem solar cells has increased to 29.8%^[1], which has broken the theoretical efficiency of a single junction silicon cells. With rapid progress in increasing PCE, perovskite/silicon tandem devices have become an inventive competitor in the photovoltaic performance race^[2-4]. Even though the single-junction perovskite device with N–I–P architecture presented power conversion efficiencies (PCEs) of 25.5%, in tandem devices the highest performance was obtained with inverted perovskite solar cells (IPSCs)^[5-7]. Therefore, surmounting the performance loss and instability issues of inverted perovskite devices are extremely important to accelerate the commercialization of perovskite/silicon tandem technology^[8].

Up to now, the PCE of IPSCs has reached to 23.37% with a certificate efficiency of 22.75%^[9], which was the highest value reported for IPVSCs. Compared with the normal N–I–P type PSCs, the large open-circuit voltage (V_OC) deficit is considered as one of the main obstacles, particularly in wide bandgap IPSCs, which slows its performance development^{[10, 11]}. Actually, plenty of research has been done for suppressing theV_OC loss in IPSCs. These mainly include components regulation^[12-14], crystallization control^[15-17], bulk or surface defect passivation^{[11, 18-23]} and energy level alignment^[24-27]. Due to the easy solution coating process, the good physical and electronic contacts with the perovskite layer and the favorable passivation functionality, [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) is one of the popular materials for the electron transport layer in IPSCs^[28-30]. However, the direct contact between PCBM and the top metal electrode will lead to serious non-radiative recombination, thus resulting in the failing collection of charge carriers to the cathode. Bathocuproine (BCP) is the most common material used in IPSCs to block hole transport and facilitate electron collection^{[9, 18, 28-32]}. Nevertheless, this insulating layer needs to be sufficiently thin to allow electron tunneling. In general, solution-processed perovskites have a rough surface, which makes the thickness of the insulator layer difficult to control. Therefore, a buffer layer that can achieve uniform coverage on the rough perovskite surface is the most pressing need for highly efficient IPSCs.

Besides the PCE, the main challenge for actual commercialization in PSCs lies instability^[33-35]. The electron transport layers (ETLs) play a more important role in IPSCs towards stable devices. Through rational design, the ETL can provide a self-encapsulating function. For instance, You et al. employed solution-processed zinc oxide (ZnO) nanoparticles as electron transport layers, comparing with devices with organic charge transport layers, which showed ~90% retention of its initial PCE after 60 days storage in ambient^[36]. Fang et al. designed a PCBM/cerium oxide (CeO_x) nanocrystal bilayer structure with good long-term stability under ambient air storage^[37]. Tin dioxide (SnO₂) with high electron mobility and a well band alignment with the perovskite absorption layer, which is a superior material in single-junction planar PSCs or highly efficient perovskite/silicon tandem solar cells^[38-40]. Zhu et al. exploited the solution-processed SnO₂ nanocrystals (NCs) to act as efficient ETLs in the IPSCs, a PCE of 18.8% with remarkable ambient stability was obtained^[41]. Raninga et al. used atmospheric pressure chemical vapor deposition (CVD) to achieve a pinhole-free and conformal SnO₂ layer, and thus protect the perovskite against mechanical and environmental destruction, resulting in a PCE of 19.4% in an IPSCs^[42].

Here, SnO₂ using the atomic layer deposition (ALD) method was designed as an additional buffer layer on the BCP layer for improving efficiency and stability in wide bandgap IPSCs. Compared with previous methods, ultra-thin and compact SnO₂ could be obtained by ALD process. Moreover, vacuum conditions and suitable for large area fabrication also make it great potential for future commercialization. It was found that SnO₂ layer was homogeneously coated on the sample surface eliminating the possible leakage current and thus raising the shunt resistance (R_sh) and reducing the reverse current (J₀). Finally, 21.13% efficiency was obtained in the wide-bandgap (1.63 eV) inverted perovskite device with a SnO₂ additional buffer layer. Moreover, ALD growing SnO₂ layer with a great compactness which enhanced the moisture contact angle and improved stability of the perovskite device.

ITO substrates were cleaned by detergent, DI water, acetone and isopropanol for 30 min, respectively. Then the substrates were treated by UV-ozone for 20 min. And then, PTAA (5 mg/mL in CB) solution was spin-coated onto the substrates at 5000 rpm for 30 s and annealed at 100 °C for 10 min in the glovebox with nitrogen condition. 1.5 M PbI₂ : PbBr₂ (83 : 17) precursor solution was spin-coated onto PTAA at the speed of 3000 rpm for 30 s and annealed at 70 °C for 1 min. And then, a mixture solution of organic salt (FAI : MABr : MACl = 90 : 6 : 9) was spin-coated onto the film at a speed of 3500 rpm for 30 s followed by annealing at 150 °C for 20 min in ambient air (30%–40% humidity). The passivation layer GABr (3 mg/mL in IPA) was fabricated by spin-coating at 5000 rpm for 30 s followed by annealing at 100 °C for 5 min. Subsequently, PCBM solution (20 mg/mL in CB) was spin-coated onto the perovskite film at a speed of 1500 rpm for 30 s. BCP solution (0.5 mg/mL in IPA) was spin-coated onto the PCBM film at a speed of 5000 rpm for 30 s. And then, the devices were moved into ALD devices to deposit SnO₂ film with different thicknesses. The deposition of the SnO₂ with the following processing sequence: 0.5 s TDMASn pulse, 15 s purge (20 sccm N₂), 0.1 s deionized water pulse, and 15 s purge (20 sccm N₂) at 85 °C. Finally, 150 nm of Cu top electrode were thermally evaporated under vacuum with a shadow mask. The area of the cells is 0.12 cm², when measuring, a 0.0887 cm² mask was used to define the accurate aperture area for the cells.

Lead bromide (PbBr₂), chlorobenzene (CB), dimethylsulfoxide (DMSO), N, N-dimethylformamide (DMF) and isopropyl alcohol (IPA) were purchased from Sigma-Aldrich. Polybis (4-phenyl) (2,4, 6-trimethyl-phenyl) amine (PTAA), [6,6]-Phenyl C61 butyric acid methyl ester (PCBM), bathocuproine (BCP), lead iodide (PbI₂), bromide guanidine (GABr), formamidinium iodide (FAI), methylammonium bromide (MABr), and methylammonium chloride (MACl) were purchased from Xi’an Polymer Light Technology Corp.

Scanning electron microscopy (SEM, FEI NanoSEM650) was used to characterize the morphology of the SnO₂ films and also the cross-sectional image of the fully device. The current density–voltage (J–V) characteristics of the devices were measured by Keithley 2400 Source meter under AM 1.5G illumination with Xenon-lamp based solar simulator (Enli. Tec., Taiwan). The intensity of the light was calibrated with KG-5 filter Si photodiode. And the area of the solar cells was 0.08875 cm². When measuring the device performance, a metal mask was used to reduce the influence of the scattered light. The forward scan (0–1.2 V, scan rate 20 mV/s, delay time 20 ms) and reverse scan (1.2–0 V, scan rate 20 mV/s, delay time 20 ms) were applied. External quantum efficiency (EQE) spectra were conducted by Enli Tec (Taiwan) measurement system in the wavelength from 300 to 900 nm with a scanning step of 10 nm. The photoluminescence (PL) and time-resolved photoluminescence (TRPL) spectroscopies were measured with a laser wavelength of 475 nm and power of 0.2 mW (Edinburgh FS5). The root mean square (RMS) roughness of different films was taken from atomic force microscopy (AFM) images measured with ARTCAM-130-SIINT (ARTRAY Co. Ltd., Japan).

Firstly, SnO₂ with different thickness as an additional buffer layer is designed. Fig. 1(a) presents the illustration of the IPSCs devices with a structure of ITO/PTAA/Perovskite/PCBM/BCP/SnO₂/Cu. The thickness of perovskite film is around 550 nm (Fig. S1). Current density–voltage (J–V) curves of the devices with SnO₂ films of different thicknesses (0, 10, 20, 30, 40 nm) are shown in Fig. 1(b). The results demonstrate that the devices with SnO₂ layer obtain a higher V_OC and fill factor (FF), short-circuit current density (J_SC) is also slightly increased. The control device without SnO₂ layer achieves a PCE of 17.47%. When the thickness of SnO₂ film increased from 10–30 nm, the PCE is enhanced to 17.60, 18.29 and 20.02, respectively. While the thickness of SnO₂ increased to 40 nm, the performance of IPSCs decreased to 19.82%, which may be due to the thick SnO₂ film introducing a high series resistance. The performance of IPSCs with different scanning directions are presented in Fig. S2 and Table S1. And the statistical diagrams of the device performances are shown in Fig. S3. After optimization, the device of 30 nm SnO₂ exhibited champion performance with a V_OC of 1.19 V, J_SC of 21.86 mA/cm², FF of 81.07%, and the corresponding PCE of 21.13%. While, the control device only with a PCE of 19.23%.

J–V and external quantum efficiency (EQE) curves are presented in Figs. 1(c) and 1(d), respectively. The integral current densities of the devices are 21.41 and 21.67 mA/cm² which are well matched with the J_SC in J–V measurement. The bandgap of the perovskite can be determined by the tangent of the curve to the base line which is also presented. The ultraviolet-visible absorption of perovskite material and (αhν)² versus energy of perovskite film are shown in Fig. S4. It can be found that the bandgap of the perovskite is about 1.63 eV. Therefore, the V_OC deficit (E_g/q–V_OC) is only of 440 mV. This value is a lower V_OC loss which compares to other wide bandgap IPSCs as present in Table S2. Moreover, device performance of different SnO₂ thicknesses with or without BCP are also investigated (Fig. S5). BCP is essential when a thinner SnO₂ is deposited. With increasing the thickness of SnO₂, the role of BCP is gradually diminished. Devices with BCP and a suitable thicker SnO₂ buffer layer achieved a higher FF and PCE.

Time-resolved photoluminescence (TRPL) is employed to investigate the transport and recombination dynamics of photoinduced carriers in perovskite films without and with 30 nm SnO₂ buffer layer (Fig. 2(a)). The biexponential function is applied to fit the TRPL decay curves (Eq. (1))^[43]:

$ Y={A}_{1}{\rm e}^{-t/{\tau }_{1}}+{{A}_{2}{\rm e}}^{-t/{\tau }_{2}}+{y}_{0}, $  (1)

where τ₁ and τ₂ are the lifetimes for the carrier transport and recombination, and A₁ and A₂ are the corresponding amplitudes. τ₁ is due to the carrier transport occurring in the surface of the perovskite film, whereas τ₂ could be attributed to the bulk and surface recombination. The devices without SnO₂ obtained τ₁ = 42.01 ns and τ₂ = 146.43 ns, respectively. While τ₁ = 37.65 ns and τ₂ = 181.59 ns are given with SnO₂. The decrease of τ₁ indicates faster electron transfer from perovskite to ETL with the presence of SnO₂. The increased τ₂ indicates a slower recombination rate. The film with defferent thickness of SnO₂ film were also presented in Fig. S6, and the fitting value of the carrier transport and recombination lifetime were shown in Table S3. The results illustrate that perovskite film with SnO₂ additional layer achieves a faster carrier extraction and a lower carrier quenching between the active layer and the transport layer, which is a good agreement with the higher FF and V_OC.R_sh of the devices without or with 30 nm SnO₂ are also calculated (Fig. 2(b)). The higher R_sh benefits the enhancement of the FF, which further confirms the excellent function of the SnO₂ buffer layer.

Fig. 2(c) shows the dark J–V curves of the devices without and with a 30 nm SnO₂ buffer layer. According to the Schottky junction model, diode ideality factor (n) and J₀ can be obtained by fitting the linear J–V curve (exponential diode behavior region) as shown in Fig. 2(d). The Eq. (2) shows in the following^[44]:

$ J={J}_{0}\left[\mathrm{e}\mathrm{x}\mathrm{p}\left(\frac{eV}{nkT}-1\right)\right] , $  (2)

where J is the current density value, V is the applied voltage, T is the absolute temperature, k is the Boltzmann constant, e is the electron charge. From the equation we can find that n andJ₀ are close related with the slope and intercept of the linear fitting of ln(J₀) versus V. The control device has a slope of 16.41, while the device with 30 nm SnO₂ film has a slope of 19.38. The larger the slope, the smaller the n, which proved a better diode characteristic. The corresponding J₀ values of 1.10 × 10^–9 and 1.22 × 10^–8 mA/cm² are obtained of the devices with and without SnO₂ film. The results clearly indicated that device with SnO₂ film has a lower J₀, which means a smaller leakage current, resulting in a higher V_OC.

A top-view scanning electron microscope (SEM) is primordially verified to show the surface morphology of the devices with or without SnO₂ (Fig. 3). Figs. 3(a)–3(e) were the SnO₂ film morphologies with the thickness of 0, 10, 20, 30 and 40 nm on the structure of glass/ITO/PTAA/perovskite/PCBM/BCP, respectively. As expected, SnO₂ films with different thicknessness demonstrate different coverage, and the surface becomes increasingly compact as the thickness increases. SnO₂ with 10 nm obtains the poorest coverage while all of the devices with thickness over 20 nm show uniform and dense morphology, and the device with 30 nm SnO₂ has the best coverage. Fig. 3(f) shows a clearer morphology of SnO₂ with the thickness of 30 nm, and it further illustrates that the compact SnO₂ layer can uniform coverage of the ETL layer, obstructing the immediate contact with the metal electrode. And the illustrations present the corresponding surface morphology of the films with different thicknesses of SnO₂ by an atomic force microscope (AFM). The root-mean-square (RMS) roughness of the films is shown in Fig. 3(g), SnO₂ film deposited by ALD can effectively reduce surface roughness. The assumed schematic diagrams of the performance improvement are demonstrated in Figs. 3(h) and 3(i). Perovskites with a rough surface, so the solution method BCP with a thinner thickness may not cover the PCBM completely. This supplies an opportunity of direct contact between PCBM and metal electrode, causing a non-radiative recombination center and reducing the R_sh, and thus degrading the performance of the devices. Whereas, the ALD method SnO₂ can conformal growth on the uneven surface which completely prevents the current leakage. That more distinctly illustrates the reason of the increased V_OC and FF of the devices with SnO₂ additional buffer layers.

The water contact angles are presented in Figs. 4(a) and 4(b). It is evidently shown that the film with a 30 nm SnO₂ layer has a larger contact angle which can better resistance to moisture and then stabilize the device. The devices without and with 30 nm SnO₂ are stored in a glove box with nitrogen (N₂) condition (Fig. S7). Both of the devices show relatively high stability, the device performance with SnO₂ is basically keeping the same efficiency without decay after 600 h. Figs. 4(c)–4(e) show the normalized V_OC, J_SC and FF of the device with and without SnO₂ layer in the air atmosphere with relative humidity between 20%–40% without encapsulation, respectively. Fig. 4(f) presents the normalized PCE of the devices with or without SnO₂ layer. It is clearly shown that the device with SnO₂ still maintains 90% of initial efficiency after 600 h. Nevertheless, the efficiency of the device without SnO₂ drops down to less than its 70%. We can find that the attenuation of the PCE is mainly due to the decrease of FF and V_OC, while the J_SC are barely changed. Fig. S8 presents the changes of series resistance (R_s) for the two devices, we supposed the increased R_s is the reason for the FF and V_OC losses.

In summary, we have demonstrated that the ALD-processed SnO₂ as an additional buffer layer in IPSCs is a general approach to enhance the device’s performance and stability. The results show that this additional buffer layer can significantly increase the R_sh and reduce the leakage current, thus enhancing the FF and V_OC. Meanwhile, the uniform SnO₂ film with good hydrophobicity to repelling moisture, which can retard the decomposition of perovskite film. The device with a SnO₂ buffer layer achieves over 90% of its initial efficiency after 600 h under an ambient condition with relative humidity of 20%–40%. Instead, the control device only exhibits 70% of its initial efficiency. This work provides a facile way of boosting the efficiency and stability of IPSCs, which also gives guidance for perovskite-based tandem solar cells for future commercialization.

B.B. Chen and P.Y. Wang contributed equally to this work. The authors gratefully acknowledge the supports from National Key Research and Development Program of China (Grant No. 2018YFB1500103), the Overseas Expertise Introduction Project for Discipline Innovation of Higher Education of China (Grant No. B16027), Tianjin Science and Technology Project (Grant No. 18ZXJMTG00220), and the Fundamental Research Funds for the Central Universities, Nankai University (Grant Nos. 63191736, ZB19500204). Natural Science Foundation of Tianjin (No. 20JCQNJC02070). China Postdoctoral Science Foundation (No. 2020T130317).

Supplementary materials to this article can be found online at https://doi.org/10.1088/1674-4926/43/5/052201.