High-Performance and Hysteresis-Free Perovskite Solar Cells Based on Rare-Earth-Doped SnO2 Mesoporous Scaffold

Tin oxide (SnO2), as electron transport material to substitute titanium oxide (TiO2) in perovskite solar cells (PSCs), has aroused wide interests. However, the performance of the PSCs based on SnO2 is still hard to compete with the TiO2-based devices. Herein, a novel strategy is designed to enhance the photovoltaic performance and long-term stability of PSCs by integrating rare-earth ions Ln3+ (Sc3+, Y3+, La3+) with SnO2 nanospheres as mesoporous scaffold. The doping of Ln promotes the formation of dense and large-sized perovskite crystals, which facilitate interfacial contact of electron transport layer/perovskite layer and improve charge transport dynamics. Ln dopant optimizes the energy level of perovskite layer, reduces the charge transport resistance, and mitigates the trap state density. As a result, the optimized mesoporous PSC achieves a champion power conversion efficiency (PCE) of 20.63% without hysteresis, while the undoped PSC obtains an efficiency of 19.01%. The investigation demonstrates that the rare-earth doping is low-cost and effective method to improve the photovoltaic performance of SnO2-based PSCs.

Titanium dioxide (TiO 2 ) is a frequently used electron transport material for perovskite solar cells. However, TiO 2 shows a lower electron mobility of (0.1~1.0 cm 2 V -1 s -1 ), compared with conventional perovskite material (20~30 cm 2 V -1 s -1 ), which causes insufficient charge carrier separation at the TiO 2 /perovskite interface [23][24][25]. In addition, the UV instability of TiO 2 upon UV exposure triggers a rapid decrease in performance of PSCs via the degradation of the organic components in the PSC (especially for mesoporous TiO 2 -based PSCs) [26,27]. Furthermore, high temperature processing (HTP) for TiO 2 electron transport layer (ETL) is also unfavorable for the fabrication of low-cost PSCs. To overcome these issues, various metal oxides (i.e., SnO 2 , ZnO, WO 3 , In 2 O 3 , and SrTiO 3 ) and fullerene are investigated as the substitute electron transport materials for TiO 2 [28][29][30][31][32]. Among them, SnO 2 has emerged as an especially promising candidate, owing to its low temperature processability, high optical transmittance in visible range, high electron mobility (100~200 cm 2 V -1 s -1 ), less sensitive to UV radiation, and favorable energy level alignment to perovskite absorbers [33]. However, solution-processed planar SnO 2 films suffer from temperature-dependent nonideal electron mobility at low annealing temperature or crack defect morphology brought by high annealing temperature (i.e., 500°C), which both lead to inferior contact and electron transport at the ETL/perovskite interfaces and then manifested in J-V hysteresis [34][35][36]. Fortunately, a thin mesoporous ETL can be designed to ameliorate these issues by improving perovskite coverage condition to get large perovskite grains [37][38][39]. On this basis, mesoporous SnO 2 (m-SnO 2 ) ETL has been highlighted for the industrial production of stable and efficient SnO 2 -based PSCs [40]. The PSCs embedded with full SnO 2 blocking layer (bl)/mesoporous (mp) layer have obtained inspiring advances [41]. Higher PCEs of 13.1% and 17% were obtained employing HTP m-SnO 2 and gallium-doped m-SnO 2 by Roose et al. [40,42]. Afterwards, Liu et al. and Yang et al. reported the PSCs with LTP 2D SnO 2 nanosheet arrays and yttrium-doped SnO 2 nanosheet arrays; the PCEs increased to 16.17% and 17.29% [34,43]. Recently, Xiong et al. reported a highstabilized PCE of 19.12% on the planar Mg-SnO 2 PSC by using HTP m-SnO 2 scaffold [36]. More recently, lowtemperature processed (LTP) planar SnO 2 PSCs have been achieved a PCE of 21.4% by using tin oxide precursors based on acetylacetonate [44].
In this work, monodisperse mesoporous SnO 2 nanospheres with large surface area are synthesized under 300°C posttreatment and are used as mesoporous scaffold in PSCs to improve photovoltaic performance of devices by modifying perovskite coverage condition with large perovskite grains without damaging the surface morphology of SnO 2 blocking layer. Moreover, rare-earth cations located in the third subgroup (Sc 3+ , Y 3+ , La 3+ ) are introduced into the SnO 2 mesoporous scaffold to reduce trap state density and transport resistance, enhance charge carrier concentration, and regulate energy level of SnO 2 , resulting in enhancement of photovoltaic parameters of the device. By optimizing the amount of Ln 3+ dopants, the 3%-SNOY device exhibits a hysteresis-free and high-stabilized power conversion efficiency of 20.63%, superior to those reported previously for full SnO 2 mesoporous structure PSCs. Furthermore, the UV stability is also investigated to illustrate the excellent longterm stability of these full SnO 2 -based PSCs.

Results and Discussion
2.1. Structure and Morphologies. For research, SnO 2 were synthesized with planar (p-SNO), mesoporous (m-SNO), and doped Ln in Ln 3+ /Sn 4+ molar ratio (x%-SNOLn). Figure 1(a) shows the X-ray diffraction pattern (XRD) of m-SnO 2 and 3% Ln 3+ (Sc 3+ , Y 3+ , La 3+ )-doped m-SnO 2 . All peaks are readily indexed to the tetragonal rutile phase of SnO 2 (JCPDS card No. ; no additional peaks are observed, indicating that Ln 3+ (Sc 3+ , Y 3+ , La 3+ ) dopants do not change the phase structure of SnO 2 . The range (2θ = 24~28°) is magnified in Figure 1(b). The peaks for doped samples are slightly broader and shift to lower angels compared with the peak for undoped sample. This is due to the partially substitutional or interstitial incorporation of Ln 3+ and the radius difference between Sc 3+ (0.087 nm), Y 3+ (0.1019 nm), La 3+ (0.116 nm), and Sn 4+ (0.069 nm), suggesting there is no distortion in the bulk SnO 2 lattice due to the small amount of doping.
The X-ray photoelectron spectroscopy (XPS) scans for all samples, as shown in Figure S1a-S1c, highlight that no impurities are found in the spectra, suggesting that the Ln 3+ ions are incorporated successfully into m-SnO 2 . From the high-resolution XPS spectra of Sn 3d (Figure 1(c)), undoped m-SnO 2 shows two signal peaks at 486.7 and 495.2 eV, respectively, corresponding to Sn 3d 5/2 and Sn 3d 3/2 states of Sn 4+ . After doping, all Sn 3d peaks shift to lower binding energies, arising from the variations of the chemical environment for Sn 4+ . In detail, the different binding interactions between Ln-O and Sn-O lead to the charge transfer effect around Sn 4+ species [45]. Besides, compared with the binding energy of Sc 2 O 3 (401.9 eV), Y 2 O 3 (156.6 eV), and Ln 2 O 3 (835.2 eV) [46], the shift to lower binding energy with different degrees indicates that Ln 3+ exists in a substitutional or interstitial bonding mode of Sn-O-Ln [47]. The narrow spectra of Sc 2p, Y 3d, and La 3d are manifested in Figure 1(d) and Figure S1d, S1e; all characteristic peaks can be assigned to +3 oxidation state of Ln [48,49]. Figure 2(a) shows TEM image of 3% Y-doped m-SnO 2 , which mainly consists of monodisperse nanospheres with the average diameter about 54:19 ± 4:44 nm (Figure 2(c)). The enlarged TEM image shown in Figure 2(b) reveals that the 3% Y-doped m-SnO 2 nanospheres have granular structure and coarse surface that consist of small SnO 2 nanoparticles with a size of 6~8 nm, which endows the 3% Y-doped m-SnO 2 nanospheres with high surface area. From Figure 2(b) and Figure S2, the lattice spacing for m-SnO 2 and 3% Ln 3+ -doped m-SnO 2 is calculated as 0.334 nm, corresponding to the rutile SnO 2 phase of (110) lattice planes, indicating that the Ln 3+ dopants have no influence on the crystalline phase of m-SnO 2 , which is accorded with the XRD patterns ( Figure 1(b)).
The pore width and the specific surface area of undoped m-SnO 2 and 3% Y 3+ -doped m-SnO 2 , derived from nitrogen adsorption-desorption measurements, are shown in Figure 2(d) and Table S1. All samples exhibit a type H3 hysteresis loop according to the Brunauer-Deming-Deming-Teller (BDDT) classification, indicating the presence of mesopores (2~50 nm) [50,51]. For 3% Y 3+ -doped m-SnO 2 , the Barrett-Joyner-Halenda (BJH) desorption cumulative pore width is 13.8 nm, which is in agreement with the above BDDT classification. Compared with the pore width of undoped m-SnO 2 sample (9.80 nm), an increment of 40.8% is achieved by 3% Y 3+ dopant [34]. According to the BET method, the specific surface areas of undoped m-SnO 2 and 3% Y 3+ -doped m-SnO 2 are 120.2 and 130.0 m 2 g −1 , respectively. Based on the BET results, the pore feature (especially for pore width) of the doped sample is improved due to the addition of rare-earth ions, which is profitable for the penetration of the perovskite into the mesoporous scaffold and then the formation of wellaligned perovskite morphology. Figures S3a-S3c show the FE-SEM images of m-SnO 2 nanospheres with similar particle diameter for undoped and doped samples, indicating the doping hardly changes the morphology of particles; thus, the influence of morphology by doping on device performance can be excluded. However, the granular structure of m-SnO 2 nanospheres collapses when the Ln 3+ concentration is increased to 4% ( Figure S3d, S3e) [34,37]. Thus, a suitable Ln doping amount is crucial. In our experiment condition, the Ln optimal concentration is 3%. Figure 3 shows the FE-SEM images of SnO 2 films (annealed at 300°C), perovskite films, and the devices. Figure S3 shows the corresponding FE-SEM images annealed at 180°C. In Figure 3(a) and Figure S3g, the morphologies of the p-SNO annealed at 300°C and 180°C are similar, indicating that the p-SNO films are thermal stable. From Figure 3(b), 3%-SNOY thin film consists of many monodisperse and well-aligned nanospheres, which increases contact area between m-SnO 2 scaffold/perovskite layer and results in the improvement of J SC and FF. From    Table S2. The gradual upward shift of the E F confirms the improvement in carrier concentration crosschecked by conductivity and Mott-Schottky measurements mentioned after [43,52]. According to E VB = E F − E on−set (on-set binding energy) and E CB = E VB + E g , the valence band positions (E VB ) and the conduction band (E CB ) are calculated and expressed in Figure 4(d) and Table S2. SnO 2 mesoporous scaffold plays a functional bridging role between SnO 2 bl and perovskite layer; the increase of E CB values of p-SNO, m-SNO to SNOLn films is beneficial for the charge carrier extraction, which the results are good coincidence with the J-V measurements. Meanwhile, the enhancement of V OC can be attributed to conduction band upward shift, influenced by the E CB and E VB of ETL and hole transport layer (HTL). According to the above discussions, the enhancement in photovoltaic parameters is ascribed to the optimization of morphology of perovskite film and charge transport dynamics.  , which is inversely proportional to the straight line slope of the M-S plot using the equation: slope = 2/εε 0 A 2 qN e , where ε is the relative dielectric constant for SnO 2 , ε 0 is the vacuum permittivity, A is the sample area, and q is the elementary charge [43,54,55]. The slope of the 3%-SNOY film (2:10 × 10 16  smaller than that of m-SNO film (10:22 × 10 16 ), suggesting a considerable increment of N e . Simultaneously, it also confirms that Ln 3+ ions are incorporated interstitially into SnO 2 as p-type dopants, which make the increment of N e [43].
To further elucidate the effect of Ln 3+ doping on hysteretic behavior of the device, we carried out steady-state photoluminescence (PL) quenching experiment and timeresolved photoluminescence (TRPL) intensity decay measurement. As presented in Figure S4c, the samples with mesoporous scaffold show a faster electron quenching efficiency than the p-SNO film. According to the result, 3%-SNOY shows the fastest quenching rate, indicating an efficient electron transport and charge extraction at the interface of perovskite/SnO 2 , which can prevent the accumulation of redundant capacitive charge that leads to hysteresis. Figure S4d and Table S3 show the TRPL decays of samples; the curves were fitted with a twocomponent exponential decay function I = A 1 e −ðt−t0Þ/τ1 + A 2 e −ðt−t0Þ/τ2 , where τ 1 refers to the faster component of trap-mediated nonradiative recombination and τ 2 is the slower component correlated to radiative recombination. The average PL decay times (τ ave ) can be calculated with the formula τ ave = ðA 1 τ 1 2 + A 2 τ 2 2 Þ/ðA 1 τ 1 + A 2 τ 2 Þ, in which A 1 and A 2 represent the decay amplitudes. The fitted result for the 3%-SNOY film delivered a much faster PL decay rate (58.89 ns) than that of the p-SNO film (171.35 ns), validating the trap-assisted recombination is partly eliminated, which is beneficial for improving PCE output. Figure S4e illustrates the Nyquist plots of PSCs devices; the intercept point on the real axis in the high-frequency range is the series resistance (R s ), the semicircle in the    high-frequency range is the charge-transfer resistance (R ct ) between at HTL/Au interface, and the value in lowfrequency range is the recombination resistance (R rec ) at mesoporous/perovskite layer interface. The measured resistances are listed in Table S4, the R rec trend is in good coincidence with the J-V measurements, and the improvement of FF can be attributed to the optimization of charge-transfer resistance of devices owing to Ln 3+ doping. Figure 5(a) shows the structure diagram of fabricated PSCs. Figure 5(b) shows the characteristic photocurrent density-voltage curves (J-V) curves of the optimized p-SNO, m-SNO, and 3%-SNOLn (Sc 3+ , Y 3+ , La 3+ ) devices; the relative photovoltaic data are given in Table S4. Impressively, compared to the p-SNO device, m-SNO cell shows an improved PCE of 19.01%, indicating a 10.46% enhancement in PCE (19.01% vs. 17.21%), which attributes to efficient electron transport brought by dense, large size, and vertical distribution of perovskite (Figure 3(h)) and improved interfacial connection between SnO 2 scaffold and perovskite. We note that PCE reach a highest value of 20.63% for 3%-SNOY device, owing to the morphology optimization of perovskite layer, the adjustment of energy level alignment with perovskite layer, and the improvement of charge transport dynamics by Y 3+ doping. The performance of PSCs is further improved owing to Ln ion doping, obeying an order of 3% − SNOY > 3% − SNOSc > 3% − SNOLa > m − SNO > p − SNO, yielding a champion PCE as high as 20.63% for 3%-SNOY tailored PSC. The photovoltaic performance of the 3%-SNOY device is higher than that of state-of-the-art PSCs based on mesoporous SnO 2 scaffold (Table S5). Figures S5a-S5c and Table S6 show J-V curves and photovoltaic parameters of the PSCs based on m-SNO layer doped with different Ln ion concentrations. To ensure the reliability and repeatability of data, average photovoltaic parameters of the each PSC device were obtained from 20 devices. As mentioned above, excessive dopants are detrimental to the device performance either by the hoist of conduction band leading to an inefficient electron injection, in which case V OC keeps increasing, but J SC decreases, or the dopants induce trap states, leading to a decrease of all device parameters [43]. In our case, both the trap state density increase induced by heavily Ln 3+ -doped m-SnO 2 (4%-Sc 3+ , Y 3+ , La 3+ ) and the dreadful recombination occurring at the interface of mesoporous scaffold/perovskite induced by the scaffold breakdown drag the parameters of PSC devices. The reproducibility statistics of 20 devices shown in Figures S5d-S5g and Table S7 further confirm the better photovoltaic data for the 3%-SNOLn devices than the p-SNO devices. Apparently, the 3%-SNOY device exhibits the highest average PCE value (20:26 ± 0:116%), and average J SC , V OC , and FF values of 3%-SNOY device are the highest among all devices. The reliability and repeatability tests distinctly demonstrate the champion performance of the 3%-SNOY devices. Figure 5(c) shows the steady-state output of J SC and PCE for m-SNO and 3%-SNOY devices by tracking maximum power point (MPP) at a bias voltage 0.87 V and 0.94 V as shown in Figure 5(b). The p-SNO and 3%-SNOY devices yield stabilized PCE of 16.79% and 20.28%, respectively, which are comparable to the PCE obtained from the fresh J-V curves. Figure 5(d) demonstrates incident photon-tocurrent conversion efficiency (IPCE) spectra of p-SNO, m-SNO, and 3%-SNOY champion devices to verify the validity of the power output. The 3%-SNOY device exhibits the highest integrated J SC value of 22.75 mA cm −2 compared to m-SNO (22.18 mA cm −2 ) and p-SNO (21.15 mA cm −2 ), concluding an efficient charge injection and well energy level alignment for 3%-SNOY film. In our case, the J SC values derived from IPCE spectra are undervalued compared with that obtained from the corresponding J-V curves ( Figure 5(b)), which is acceptable due to the spectral mismatch of the solar simulator and the theoretical AM 1.5G spectrum as reported before [56].  Table S8. Apparently, p-SNO device shows a mild hysteresis while 3%-SNOY device embodies a character of hysteresis-free. It denotes that the introduction of rareearth ions improves charge carrier transport dynamics and suppresses the charge accumulation at the interfaces of SnO 2 bl/m-SnO 2 and m-SnO 2 /perovskite layers. Figures S6a and S6b show the dependence of J SC and V OC on light intensity of the PSCs. The power law dependence of J SC on the illumination intensity is generally expressed as J SC ∝ I α , where α is the exponential factor related to bimolecular recombination [56]. The α value of 3%-SNOY (0.983) device is closer to 1 than that of m-SNO (0.973) and p-SNO (0.953) devices, suggesting a reduction of the bimolecular radiative recombination in the Ln 3+ iondoped devices due to more effective carrier transportation through the interfaces of SnO 2 bl/m-SnO 2 /perovskite layers.

Photovoltaic Performance.
As shown in Figure S6b, the trap-assisted Shockley-Read-Hall recombination associated with trap state density is significantly suppressed due to the Ln 3+ dopants determined as V OC = nkTlnðIÞ/q + constant [54], where n is an ideal factor related to monomolecular recombination, k represents the Boltzmann constant, and T is the absolute temperature. The slope of the plot (n) for the 3%-SNOY device is lower (1.32 kT/q) than that of the m-SNO (1.48 kT/q) and p-SNO devices (1.86 kT/q), indicating that the fewer monomolecular recombination in the 3%-SNOY device by Ln 3+ doping, which is consistent with the Figure S6a. Figure S6c shows the dark J-V curves of PSCs, which is correlated to the leakage current from carrier recombination in the devices. According to the equation, V OC = nkTlnðJ SC /J 0 Þ/q, where J SC and J 0 are the photogenerated and dark saturation current densities, respectively [57]. The value of J 0 for the 3%-SNOY device is calculated as 8:41 × 10 −11 mA cm −2 , which is five orders of magnitude smaller than that of the p-SNO device (1:32 × 10 −6 mA cm −2 ). While all three devices exhibit approximate output current, implying a higher rectification ratio and a greatly restrained leakage current induced by charge recombination for the device Ln 3+ doped. Figure S6d illustrates the dark J-V curves of the electron-only devices with a structure of FTO/SnO 2 /perovskite/PCBM/Au. Apparently, the trap-filled limit voltages (V TFL ) for 3%-SNOY device show a decreased trend in comparison with the other two devices, which is proportional to trap state density (n trap ) according to V TFL = qn trap L 2 /2εε 0 , where L is the thickness of the electron-only device, ε is the relative dielectric constant for SnO 2 , and ε 0 is the vacuum permittivity [58]. These results provide strong evidences that the trap defects are healed significantly owing to Ln 3+ ion doping, crosschecking the aforementioned conclusion.
2.5. Stability. The long-term stability of perovskite solar cells should consider both ultraviolet light and humidity sensitivities [39,59]. Figure 6(a) shows the normalized PCE vs. time of the devices without encapsulation under simulated solar light illumination for 400 h in 30% RH ambient condition. The m-TiO 2 PSCs manifest a dramatic degradation to 64.8% of the highest PCE in the first 40 h arising from the photocatalytic properties of TiO 2 degrading perovskite and then followed by a plain to 53.2% of the highest PCE originating from that the defects in TiO 2 can be passivated by the adsorption of atmospheric oxygen [39]. While the 3%-SNOY devices decrease steadily at nearly 75.8% of the highest PCE under the same testing conditions, exhibiting remarkablely improved UV stability. Table S9 shows the change of average PCE values of 3%-SNOY and m-TiO 2 devices with time, presenting similar results as Figure 6(a).
After exposure of the perovskite based on m-TiO 2 to AM1.5 illumination under 30% RH ambient condition for one week shown in Figure 6(b), a new sharp diffraction peak at 12.7°corresponding to the PbI 2 (001) lattice plane is observed in the XRD pattern, arising from the decomposition of the aged perovskite film caused by UV light and ambient humidity. The PbI 2 would block the charge transport, resulting in a decrease of PCE. Notably, the intensity of PbI 2 diffraction peak for the 3%-SNOYbased perovskite film aged in the same situations is very less, implying a much lower degree of decomposition, which is due to the passivation effect on defects and UV light stability by rare-earth ion doping.

Conclusions
In conclusion, rare-earth Ln-doped monodisperse SnO 2 nanospheres with specific surface area of 130.0 m 2 g −1 are successfully synthesized by a solution-phase route. The doped SnO 2 nanospheres are used as scaffold to fabricate mesoporous perovskite solar cells. The observation of morphology, microstructure characterization, energy band analysis, photoelectric property investigation, and photovoltaic performance measurement indicate that the doping of rareearth Ln ions promotes the formation of dense, even and large perovskite crystals, which facilitate better interfacial contacts of electron transport layer/perovskite layer. On the other hand, Ln dopants optimize the energy level of electron transport layer and reduce the resistance and charge trap states, resulting in an efficient electron transport and charge extraction. As a result, the Y 3+ (3%)-doped mesoporous SnO 2 -based PSC achieved a champion efficiency of 20.63% with hysteresis-free, while the planar and mesoporous SnO 2 -based PSCs obtain efficiency of 17.21% and 19.01%, respectively. This investigation demonstrates a novel strategy for developing efficient and low-cost full SnO 2 -based PSCs. , and ethylene glycol (EG, 99%) were obtained from Aladdin. All the chemicals were used as received without further purification. Deionized water (resistivity > 18 MΩ) was obtained through a Millipore water purification system. Prepatterned fluorine-doped tin oxide-coated (FTO) substrates with a sheet resistance of 14 Ω sq −1 were purchased from Pilkington.

Preparation of SnO 2 and
Ln-Doped Solutions. SnO 2 solution was prepared by dissolving SnCl 2 ·2H 2 O in 1-butanol in the concentration of 0.1 M. Monodisperse SnO 2 nanospheres were synthesized by a solution-phase route [50]. In a typical procedure, 8 mL of aqueous solution containing LnCl 3 (Sc 3+ , Y 3+ , La 3+ ) and K 2 SnO 3 with different Ln 3+ /Sn 4+ molar ratios (0~4%) mixed with 15 mL of EG was added into a 50 mL Teflon-lined autoclave and maintained at 170°C for 13 h. The air-cooled precipitation was washed thoroughly with deionized water for removal of K + and organic residue followed by a centrifugation and then diluted with deionized water in the concentration of 0.1 g mL −1 prior to use.

Fabrication of Perovskite Solar Cells.
Laser-patterned FTO glass with size of 1:5 × 1:5 cm 2 was cleaned by detergent and sonicating in isopropanol, acetone, deionized water, and ethanol and finally treated with UV ozone for 30 min. SnO 2 block layer (bl) was deposited on FTO by a spin-coating step (3000 rpm, 30 s) and then annealed at 150°C for 1 h, named as p-SNO. SnO 2 scaffold layer contained different Ln 3+ /Sn 4+ molar ratios (0~4%) with thickness about 100~200 nm was covered by spin-coating the SnO 2 solution at 2000 rpm for 30 s and then annealed at 300°C for 1 h to remove the organic residue, named, respectively, as m-SNO and x%-SNOLn (x = 1, 2, 3, 4). The cesium-containing triple cation perovskite was deposited by an antisolvent method according to the literatures [27,60]. The perovskite precursor solution was spin-coated at 1000 rpm for 10 s and then at 6000 rpm for 20 s on the p-SNO and full SnO 2 as bl/mp layers substrates. During the second step, 130 μL of chlorobenzene was poured on the spinning substrate 5 s prior to the end of the program to rinse out residual DMSO and DMF in the precursor films. Afterwards, the substrates were heated immediately at 100°C for 1 h and then were cooled down to room temperature naturally. Subsequently, the spiro-OMeTAD layers were subsequently deposited on top of the as-prepared perovskite layers by spin-coating 20 μL of chlorobenzene solution containing chlorobenzene (1 mL . Average photovoltaic parameters of the PSC devices were obtained from 20 devices to ensure the reliability and repeatability of data. Dark J-V curves were measured on a Keithley 2420 source meter in the dark. The stabilized power output was recorded close to the maximum power point, which was extracted from the J-V curves on an electrochemical work station (CHI660E, Chenhua Co. Ltd, Shanghai) under simulated sunlight irradiation with intensity of 100 mW cm −2 at AM 1.5G. The incident photo-to-current conversion efficiency (IPCE) curves were measured as a function of wavelength from 300 nm to 800 nm using a QE-R quantum efficiency measurement system (Enli Technology Co. Ltd). The electrochemical impedance spectroscopy (EIS) measurements were conducted on a Zennium electrochemical workstation (IM6) under AM 1.5G with the frequencies from 100 mHz to 1 MHz, the bias of 0 V, and the amplitude of 20 mV. Long-term stability under persistent moisture (30% RH) was tested by XRD measurement for 7 days and record of time-dependent photovoltaic performances for 400 h under ambient condition with 30% RH. All the average values for long-term stability were obtained from 8 devices for each sample.

Conflicts of Interest
The authors declare no conflict of interest.  Table  S1: pore properties of 3% Y-doped m-SnO 2 and undoped m-SnO 2 . Table S2: band edge (Eg), Fermi level (EF), valence band (EVB), and conduction band (ECB) of samples. Table S3: fitted data from TRPL spectra in Figure  S4d. Table S4: photovoltaic and impedance data of the PSCs. Table S5: PCE comparison of the PSCs based on full SnO 2 mesoporous structure [S1-S8]. Table S6: photovoltaic data of the PSCs with different Ln3+ (Sc3+, Y3+, La3+) concentrations. Table S7: average photovoltaic data of the PSCs. The average values were obtained from 20 devices. Table S8: photovoltaic parameters of the PSCs scanning in different directions.