The past few years have witnessed the significant development of organic-inorganic halide perovskite photovoltaics. The highest certified power conversion efficiency (PCE) of perovskite solar cells (PSCs) has reached up to 25.2% until now (NREL, 2019), which is close to that of the commercial polycrystalline silicon arising from high charge carrier mobility, long carrier diffusion length, and high light absorptivity across whole visible region (Green et al., 2014; Huang et al., 2017; Yang et al., 2018). Unfortunately, organic-inorganic halide perovskites suffer from intrinsicly thermal instability and easy-degradation under oxygen and water, which could be overcome by employing the all-inorganic halide perovskite CsPbX₃ (X=Cl, Br, I) without volatile organic components (Cho et al., 2018). It is worth mentioning that the stability of perovskite solar cells is determined by the stability of perovskite materials to some extent, so the performance degradation is unavoidable using unstable perovskite composition. Bulk CsPbI₃ with cubic structure has a small bandgap of 1.73 eV, which is ideal sunlight absorbing material. However, it undergoes immediate phase transition from cubic to orthorhombic phase when exposed to ambient condition which is due to the orthorhombic yellow phase is thermodynamically preferred (Swarnkar et al., 2016; Sanehira et al., 2017; Zhang et al., 2017; Li B. et al., 2018; Wang et al., 2019). The large bandgap of CsPbCl₃ makes it impossible in photovoltaic applications. CsPbBr₃ has a direct gap of 2.36 eV, as well as high phase stability, thermal stability, and photo-stability. Although the overall PCE is limited by a relatively larger bandgap, CsPbBr₃ has been targeted as a potential material for stable high-voltage perovskite solar cells (Akkerman et al., 2016; Liang et al., 2016; Hoffman et al., 2017; Duan et al., 2018). More importantly, the all-inorganic PSCs using CsSnBr₃ QDs as interfacial decoration produce a maximized PCE of 10.60%, which is the highest reported PCE of CsPbBr₃ PSCs to date (Zhao et al., 2019). Compared with PSCs based on pure CsPbBr₃ nanocrystals (NCs), ion-doped or mixed PSCs have also attracted increased interests. CsPbBrI₂ has reasonable broad bandgap, and stable cubic phase at room temperature, with a continuously improved PCE (Sutton et al., 2016; Niezgoda et al., 2017; Zeng et al., 2018; Zhang J. et al., 2018). It was reported that Cs_0.91Rb_0.09PbBr₃ tailored solar cell achieved a high PCE of 9.86% resulted from the suppressed non-radiative losses and radiative recombination (Li Y. et al., 2018). The CsPbI_3−xBr_x solar cell reported by Ye et al. achieved an open circuit voltage of 1.25 V and a highest PCE of 18.64% for all-inorganic PSCs. By inserting LiF in SnO₂ layer, the passivation of interface defects and optimized band alignement are achieved (Ye et al., 2019).

The fabrication technique has a significant impact on the film morphology and final device performance. Both one-step method and two-step method have been successfully employed to fabricate organic-inorganic halide PSCs with a high PCE exceeding 20% (Jeon et al., 2014; Seok et al., 2018). However, the solution-based deposition methods are difficult to directly adapt for CsPbBr₃ because of the low solubility of cesium precursors in commonly used solvents (Kulbak et al., 2015; Chang et al., 2016; Hoffman et al., 2017). Besides, both methods are sensitive to processing conditions because crystal quality, film morphology, and optoelectronic performance are controlled in the same optimization step (Akkerman et al., 2016). Perovskite NCs provide an option to solve the above problems, and the crystal structure, as well as the optical/electrical properties were summarized and discussed in detail (Deng et al., 2019). The perovskite NCs are synthesized at the high temperature to guarantee the excellent crystal quality. The unique luminous properties, including a high photoluminescence quantum yield (PLQY) and a narrow PL emission line width are related to the low defect density and reduced non-radiative recombination process (Song et al., 2015, 2016), which is beneficial to obtain PSC with a high PCE. Moreover, perovskite NCs with a large surface/volume ratio have been demonstrated to restrain the generation of yellow phase and improve air stability (Swarnkar et al., 2016; Cho et al., 2018). Despite these advantages, a series of post-treatment techniques including purification and anneal temperature need to be optimized to fabricate compact and thickness-controllable perovskite NC films by layer-by-layer deposition method.

In this work, the morphology and optical spectra of CsPbBr₃ NCs are characterized. The purification and annealing temperature was effectively adjusted to obtain uniform and compact NC films with the purpose of improving conductivity. X-ray diffraction (XRD), PL spectra and transmission spectra were employed to study the influence of purification time on the crystal and the surface property of NC films. The impact of annealing temperature and film thickness on photovoltaic performance were investigated. Finally, the current density-voltage curves under different sweep rates and external quantum efficiency (EQE) were obtained on the device with the highest PCE, and its stability was also evaluated.

The TEM images of CsPbBr₃ NCs are shown in Figure 1A. Uniform nanocubes are clearly observed with an interplanar spacing of 0.58 nm, which is consistent with the (200) plane in cubic phase (Protesescu et al., 2015; Zhang M. et al., 2018). Figure 1B shows a narrow size distribution with an average size of 19.47 nm for CsPbBr₃ NCs, which is in accordance with the narrow PL full width at the half maximum (FWHM) of 110 meV in Figure 1C. CsPbBr₃ has a Bohr exciton diameter of 7 nm and an exciton binding energy of 40 meV, which are responsible for a sharp excitonic absorption peak at 511 nm showed in Figure 1C. The PL peak at 525 nm indicates that no obvious quantum size effect is observed because the size of NCs is larger than Bohr exciton diameter, and a small Stokes shift of about 70 meV is obtained. CsPbBr₃ has a direct bandgap and the determined bandgap from the Tauc plot in Figure 1D is 2.30 eV, which is larger than that of bulk (2.25 eV) but a little smaller than that of nanosheet (2.32 eV) (Stoumpos et al., 2013; Song et al., 2016).

The colloid CsPbBr₃ perovskite NCs are stably dispersed in non-polar solvents because of the hydrophobic oleic acid and oleylamine ligands. In order to purify NCs, polar solvents are added as antisolvents to achieve centrifugation separation. Unfortunately, perovskite NCs are unstable (prone to dissolution) in polar solvents, resulting from their intrinsic ionic nature. As a result, antisolvent species and amounts, centrifugation speeds, and purification times are optimized to obtain well-dispersed NCs with high concentrations in order to meet the requirement of spin-coating NC films (Kim et al., 2015). Ethyl acetate was demonstrated to have the minimal side effect as an antisolvent (Li et al., 2016), but amounts and purification times are varied from batch to batch. The effects of purification times on the morphology of NC films are shown in Figures 2A–C. A large area of the uncovered substrate is observed for one-time-washed film, and NCs are unclear because of lots of residual organic ligands. Increased film coverage and dark yellow are obvious with the increase of purification times. For three-time-washed film, a high resolution of NCs suggests a high conductivity, and compact film is the prerequisite of reducing leakage current in solar cells. XRD patterns in Figure 2D show CsPbBr₃ NCs still maintain cubic phase after different purification times. The PL peak intensity is obviously decreased along with a slight red-shift of PL peak (Figure 2E), indicating an increased NCs electronic coupling led by the removal of organic ligands. Moreover, transmittance spectra are used to verify the film thickness, as given in Figure 2F. With the increase of spin-coating time, the film thickness is found to increase for 3-time-washed film than 1-time-washed film because NC films can be redissolved during the layer-by-layer spin-coating deposition resulting from residual organic species for 1-time-washed NCs.

Figures 3A–C exhibits the surface morphology images of three NC films annealed at different temperatures, showing increased CsPbBr₃ crystal sizes with increasing anneal temperature. The XRD patterns in Figure 3D show that CsPbBr₃ NC maintains the cubic phase even after high-temperature annealing, suggesting no phase change during air-annealing. Figures 3E,F shows anneal temperature-dependent photovoltaic performance, which is mainly caused by different J_SC. The maximum PCE obtained for NC films annealed at 120°C results from its largest J_SC, which is a result of increasing carrier mobility due to decreased amount of grain boundary, in accordance with Zeng et al.'s results (Zeng et al., 2018). Although temperature annealing is beneficial to increase grain size and improve carrier mobility, the J_SC decreases with increasing temperature to 150°C. It is strange that increased grain size does not contribute to an improved conductivity. It may result from that the partial oxidization of surface can improve the contact resistance of grains. Therefore, an inert atmosphere annealing may be helpful to improve conductivity under the high-temperature annealing.

Another parameter spin-coating time has a significant effect on NC film thickness and final solar cell performance. The photovoltaic performance of CsPbBr₃ NC films is shown in Figure 4A. Both monotonically increased V_oc and J_sc are found with the increasing spin-coating times. The device using 1-spin-coated NC film failed because low film coverage leads to current leakage. The V_oc of 1.45 V for 7-spin-coated NC film is quite close to the highest report of 1.5 V for perovskite solar cells (Akkerman et al., 2016). The E_loss can be described by the equation: E_loss = E_g – eV_OC, and a small E_loss of 0.85 eV is obtained, resulting from the reduced non-radiative energy loss channels and defect states (Zeng et al., 2018). The J_sc of 4.5 mA cm⁻² is not high enough compared with the highest values of 5.6 mA cm⁻² for CsPbBr₃ NC film solar cell and 7.5 mA cm⁻² for CsPbBr₃ solar cell by one-step method (Akkerman et al., 2016; Liang et al., 2016), and low J_sc is attributed to low NC film conductivity. The relationship among spin-coating times, film thickness and device performance are summarized in Figure 4B and a linearly increased film thickness are observed. The highest PCE is obtained from 7-spin-coated NC film with a thickness of 600 nm. The increased PCE with increasing spin-coating times is attributed to increased photo-generated excitons separation due to enhanced visible light absorption and reduced current leakage.

A typical CsPbBr₃ NC film solar cell structure is FTO/TiO₂/CsPbBr₃ NC film/Spiro-OMeTAD/Au, and the thickness of every layer can be seen in Figure 5A. The best-performing solar cell exhibits a high PCE of 4.57% in Figure 5B, and we compare reported all-inorganic perovskite solar cells in Table 1. The smaller J_SC resulting from narrow visible light absorption is the reason for lower PCE compared with CsPbI₃ and CsPbBrI₂ solar cells, but their large bandgap contributes to larger V_OC. We can find increased J_SC with increased scan rate in Figure 5C, according to previous results. The EQE spectra of the device is shown in Figure 5D, and the cut-off wavelength is about 550 nm resulting from wide bandgap of CsPbBr₃, explaining its smaller J_SC compared with CsPbI₃ and MAPbI₃ solar cells (Swarnkar et al., 2016; Sanehira et al., 2017; Zhang et al., 2017; Li B. et al., 2018). The interfacial charge transfer process of device under illumination is revealed by the EIS spectra as given in Figure 5E. It is clearly observed decreased charge transfer resistance (R_ct) with the increase of applied bias voltage, which is resulted from the efficient extraction of more carriers. At the same time, the charge recombination resistance (R_r) is decreased by significantly enhancement of the free charge carriers in perovskite layer (Yang et al., 2015). The low R_ct and R_r under illumination illustrate the ideal contact interface between perovskite and transport layer. Finally, the unencapsulated device exhibits excellent stability in Figure 5F. It retains around 90% of initial PCE after 40 days of storage in air, and a reduced PCE is mainly caused by decreased J_SC, which is possibly attributed to the damage of the interface between perovskite and organic HTL. Therefore, the high air-stability contributes to solidifying CsPbBr₃ NC film solar cell as a kind of high-V_OC and stable device.

In summary, we have fabricated planar structure CsPbBr₃ NCs solar cells, and the average size of colloid CsPbBr₃ NC is 19.47 nm. The effects of purification times, anneal temperature and spin-coating times on morphology and crystal of CsPbBr₃ NC films were investigated. The compact and high conductivity NC films were prepared using a 3-time purification NC solution, and an increased NCs electronic coupling resulting from the removal of organic ligands was revealed by a slight red-shift of PL peak. We observed increased CsPbBr₃ crystal sizes with increasing annealing temperature, contributing to the increased solar cell PCE. Besides, we also found monotonically increased V_oc and J_sc with increasing spin-coating times due to increased film thickness. After the above parameters optimization, the best-performing solar cell with a high PCE of 4.57% was obtained, and the unencapsulated device exhibits great stability exceed 40 daysin air, illustrating the enormous potential of high-V_OC CsPbBr₃ NC film solar cell.

All datasets generated for this study are included in the article/Supplementary Material.

JL conceived the idea, directed, and supervised the project. LeZ took part in the device fabrication and characterization. TH, LeZ, HL, LiZ, ZL, and GW analyzed the data. All authors contributed to discussions.