Enhanced performance of planar perovskite solar cells by doping the SnO2 electron transport layer with guanidinium chloride

Tin (IV) oxide is a highly promising electron transport layer (ETL) for lead halide perovskite solar cells due to its high conductivity, transparency, wide band gap, and the possibility of low-temperature processing. Nonetheless, charge carrier recombination processes at the SnO2/perovskite interface diminish the device performance. Here, we demonstrate that SnO2 doping with guanidine hydrochloride (G-SnO2) leads to efficient surface passivation and a larger band offset between the ETL and the perovskite layer, resulting in reduced voltage losses and faster electron transfer. Moreover, G-SnO2 facilitates the growth of highly crystalline perovskite layers. Consequently, a power conversion efficiency of up to 23.48% and a high open-circuit voltage of 1.18 V are obtained in solar cells incorporating the G-SnO2 ETL. These devices also exhibited negligible hysteresis and maintained more than 96% of their initial power conversion efficiency after 1,250 h exposure to the air without encapsulation.


Introduction
The electron transport layer (ETL) is one essential part of perovskite solar cells (PSCs), assuring electron extraction from the perovskite layer and blocking holes to avoid undesired recombination processes. (Ke et al., 2015;Anaraki et al., 2016;Jiang et al., 2016). An ideal ETL should possess high electron mobility, favorable energy level alignment with the active layer, and good processability for the preparation of high-quality films. Amongst the N-type semiconductor used for the fabrication of ETLs, WO x , NbO x , CeO x , and BaSnO 3 can be found along with the more widely used materials TiO 2 , ZnO, and SnO 2 . (Murugadoss et al., 2016;Lee et al., 2017;Zhu et al., 2017). In this group, SnO 2 is a highly promising candidate owing to (i) its high electron mobility (over 420 cm 2 /V.s), favoring efficient transport; (ii) its wide energy band gap, giving rise to fewer losses due to parasitic light absorption; (iii) its low annealing temperature, normally less than 150°C, which enables the fabrication of solar cells at a lower cost and on various types of substrates including flexible ones. (Tu et al., 2019;Wang et al., 2020a;Xiong et al., 2022).
To overcome limitations originating from undesired recombination processes at the SnO 2 /perovskite interface, which are related to sub-optimal energy level alignment and interfacial defect states, (Sutanto et al., 2021;Liu et al., 2022;Odunmbaku et al., 2022), surface modification of the ETL is a commonly used approach. Various kinds of passivating agents have been reported, for example, inorganic salts, long-chain polymers, and small organic molecules. (Zhao et al., 2018;Jiang et al., 2019;Fu et al., 2020). As an example, Huang and coworkers demonstrated that passivation of the SnO 2 ETL with potassium ions (from KOH) eliminated hysteresis effects and influenced the growth of the perovskite grains, resulting in enhanced efficiency and stability. (Bu et al., 2018). Potassium chloride (KCl) has also been used for passivating SnO 2 thin films. (Jalebi et al., 2018;Wang et al., 2018;Qin et al., 2019). More recently, Yu et al. reported the use of guanidinium chloride for the treatment of the SnO 2 ETL to eliminate detrimental O 2 − species formed at its surface during thermal annealing, which led to an improvement of the power conversion efficiency (PCE) of the obtained perovskite solar cells from 15.33% to 18.46%. (Yu et al., 2020).
Compared with these post-deposition surface treatment procedures of the annealed ETL layer, in situ bulk doping of the ETL via the addition of appropriate molecular species during its preparation is an alternative approach to fine-tune its properties. (Jung et al., 2013;Ren et al., 2017;Huang et al., 2018). In particular, it can give access to increased conductivity of the ETL due to electrical doping, which is often accompanied by the positive side effect of lower annealing temperatures required for the ETL preparation. Moreover, the judicious choice of additives can lead to better control of the band alignment at the ETL/perovskite interface favoring efficient charge transfer. As an example, Li and coworkers adopted electron-donating triphenylphosphine oxide (TPPO) as a dopant of the SnO 2 layer, which enhanced its conductivity and decreased the energy barrier at perovskite/SnO 2 interface from 0.55 to 0.39 eV, resulting in a PCE of 20.69%. (Deng et al., 2020). In another example, Zheng and coworkers applied KFdoped SnO 2 to improve the performance of all-inorganic CsPbI 2 Br solar cells showing an open-circuit voltage (V oc ) of 1.31 V (Zhang et al., 2021). In our previous work, we investigated ammonium chloride as a dopant for SnO 2 , employing a double-layer ETL with different doping levels. This approach led to a significant improvement in the performance of planar PSCs due to a more favorable energy level alignment at the ETL/perovskite interface, resulting in a V oc of 1.21 V and a PCE of 21.75% (pristine ETL: 1.14 V, 18.54%). (Ye et al., 2021).
Based on these results, we explore here guanidine hydrochloride (H 2 N) 2 C = NH 2 + Cl − , also known as guanidinium chloride (Gua-Cl), as a bulk dopant for the SnO 2 ETL. Gua has been previously used mainly for improving the stability of PSCs by incorporating this large cation into the perovskite layer. (Ishibashi et al., 2017;Wu et al., 2019). In the case of methylammonium iodide (MAI)-based PSCs, the enhanced stability of the devices was attributed to the larger number of hydrogen bonds, increasing from 1-2 for MA to 6 for Gua. (Jodlowski et al., 2017). When using Gua-Cl for doping the ETL, a synergistic beneficial effect can be expected of the chloride anion, which can act as an n-type dopant in SnO 2 , and the Gua cation, enabling efficient hydrogen bonding with the halide ions from the perovskite layer. Furthermore, the delocalization of the positive charge over the three nitrogen atoms and the resulting resonance stabilization are expected to result in an improved electronic coupling with the perovskite and enhanced interfacial charge transfer. In this work, we developed an effective strategy for the preparation of guanidinium chloride-doped SnO 2 ETLs (G-SnO 2 ) at temperatures as low as 80°C, i.e., fully compatible with flexible substrates or with the integration in tandem devices. Our results reveal that the use of G-SnO 2 minimizes defects at the perovskite/ETL interface, which in turn enhances charge carrier transfer. The G-SnO 2 layer also facilitates perovskite crystal growth and offers better stability of the perovskite film by passivating undercoordinated ions. The PSCs prepared with G-SnO 2 showed a PCE up to 23.48% with negligible hysteresis and high V OC (1.18 V), while PSCs using pristine SnO 2 (P-SnO 2 ) had a maximum PCE of 20.91%. Moreover, G-SnO 2 -based devices maintained over 96% of their initial performance after being kept in the air for 1,250 h without encapsulation.

Device fabrication
ITO-coated glass substrates were cleaned in detergent solution, deionized water, acetone, ethanol and isopropyl alcohol in order before exposing them to UV-ozone for 30 min to remove residual organic materials. The SnO 2 precursor solution was prepared by diluting 1 ml of the colloidal dispersion of SnO 2 nanoparticles 6 times with deionized water. In the case of the G-SnO2 layer, 50 mmol of the dopant guanidine hydrochloride was added just before use. The obtained SnO 2 precursor solutions were spin-coated on ITO glass substrates twice with no heat treatment between the two steps using 100 μl solution, 2000 rpm acceleration, 4,000 rpm speed, and 30 s duration, and then the films were annealed at 80°C for 1 h. For the deposition of the CsFAMA perovskite film, the precursor solution was prepared by mixing PbI 2 (1.45 M), FAI (1.15 M), PbBr 2 (0.21 M), MABr (0.21 M) and MACl (0.45 M) in a mixed solvent of DMF/DMSO (8:1 v/v, 1 ml) and finally adding 17.5 μl of CsI solution (2 M in DMSO). The obtained solution was spin-coated onto the ETL substrate by two consecutive spin-coating steps at 1,000 and 4,000 rpm for 10 and 30 s, respectively. During the second spin-coating step, 150 μl of chlorobenzene were dropped onto the substrate after 20 s before transferring it to a hotplate for thermal treatment at 105°C for 60 min within a glove-box. After Frontiers in Materials frontiersin.org cooling to room temperature, 60 μl of a 20 mM PEAI solution in IPA were spin-coated onto the perovskite surface at a speed of 4,000 rpm. For the preparation of the HTL, spiro-OMeTAD was dissolved in 1 mL of chlorobenzene (80 mg/mL), with the addition of 29.8 μl of 4-tertbutylpyridine, 18 μl of Li-TFSI (520 mg/mL in acetonitrile) and 12 μl FK209 (250 mg/mL in acetonitrile) to increase the conductivity. Around 40 μl of the resulting precursor solution were spin-coated on top of the perovskite/PEAI layer at a speed of 4,000 rpm for 30 s. Finally, a 100-nm Au electrode was thermally evaporated on top of the stack.

Characterization
Morphological characterization (top-view and cross-sectional images) of the perovskite layer and full devices was performed using a high-resolution scanning electron microscope (Zeiss Ultra 55). AFM images were obtained with a Bruker Dimension Icon Atomic Force Microscope. GIXRD and XRD data were acquired on a Panalytical Empyrean diffractometer equipped with a Cobalt anode beam tube (Co λ Kα1 = 1.7890 Å, Co λ Kα2 = 1.7929 Å), a Gӧbel mirror, and a 1D Pixcel detector; to simplify the comparison with the literature, the data is presented for Cu K α radiation. UV-vis absorbance spectra were recorded on a Shimadzu UV-1800 spectrometer. Conductivity values were obtained using an Ecopia HMS-3000 Hall Measurement System, a magnetic field of 0.5T, and a current from −200 mA to +200 mA. The current-voltage characteristics of the devices were assessed with a Keithley 2,400 Source meter under AM 1.5G illumination using a calibrated solar simulator (Newport AAA). The area of the solar cells was delimited with a mask (0.09 cm 2 ) and a scan rate of 20 mVs −1 was used for both forward and reverse scans. Steadystate and time-resolved PL analyses were performed using a FLUOROLOG spectrometer (HORIBA FL3) with a 450 nm continuous or 425 nm pulsed laser excitation (NanoLED-425L). X-ray photoelectron spectroscopy (XPS) analyses were carried out with a Versa Probe II spectrometer (ULVAC-PHI) equipped with a monochromated Al Kα source (hν = 1,486.6 eV). The core level peaks were recorded with a constant pass energy of 23.3 eV. The XPS spectra were fitted with CasaXPS 2.3 software using Shirley background. Binding energies (BEs) are referenced with respect to adventitious carbon (C 1s BE = 284.8 eV). The UPS measurements were performed on an M-XPS spectrometer from Omicron Nanotechnology equipped with an ultraviolet He lamp (He I = 21.2 eV). The samples were electrically biased at −5.0 V for accurate determination of the low-kinetic energy cutoff. Photoelectrons were collected under normal geometry by a hemispherical analyzer, with an angular acceptance of 8°.

Results and discussion
We fabricated solar cell devices with n-i-p architecture comprising a stack of ITO/SnO 2 /Perovskite/Spiro/Au ( Figure 1A) using the triple-cation (cesium/formamidinium/methylammonium, CsFAMA) perovskite. The ETL was prepared by adding an optimized amount of Gua-Cl to commercial SnO 2 colloids in an aqueous solution followed by spin-coating and thermal annealing at Frontiers in Materials frontiersin.org 03 low temperature (cf. Experimental Section). Protonated Gua and chloride ions on SnO 2 promote photogenerated electron transfer from the perovskite layer because of electrostatic interactions. Due to the higher formation energy, Pb-Cl bonds are less likely to form, avoiding antisite defects and thus having little effect on interface recombination. (Grancini et al., 2015;Jiang et al., 2015;Chandrasekhar et al., 2016). To visualize the electron density distribution, electrostatic potential (ESP) analysis was performed. As shown in Figure 1B, the area of the imino moiety presents a high electron density making it prone to interaction with undercoordinated Pb 2+ cations at the perovskite grain surface, which are the main sources of charge traps. These factors are favorable for reducing hysteresis effects and improving electron extraction. Moreover, guanidinium cations can be directly incorporated into the lattice of triple cation perovskites and form a thin layer of a 2D perovskite phase between SnO 2 and the 3D perovskite, , contributing to thermal and environmental stability enhancement. XPS measurements ( Figure 1C) were used to investigate the oxidation state of tin in SnO 2 . The peaks visible at 486.3 and 494.7 eV in P-SnO 2 , corresponding to the Sn 3d 5/2 and Sn 3d 3/2 states, exhibit a shift to higher binding energy by 0.4 eV in G-SnO 2 . This shift signifies that Gua-Cl doping of SnO 2 leads to a lowering of electron density and thus a stronger oxidation of the Sn atoms, inducive of the filling of oxygen vacancies. The passivation of oxygen vacancies by the combined action of chloride ions and tetramethylammonium ions has recently been reported by Parida et al. for the case of fully inorganic CsPbI 2 Br PSCs. (Parida et al., 2021). Hall Effect measurements show that, due to the better passivation and doping, the G-SnO 2 layer exhibits an 11% decreased resistivity and a 7.5% higher electron mobility than the G-SnO 2 ETL (cf. Supplementary Table S2). Gua-Cl doping thus improves the charge extraction and transport properties of the ETL, which should be directly visible in the solar cell performances and hysteresis behavior. Figure 1D shows the J/V curves of the champion device employing the G-SnO 2 ETL, which yielded the best PCE of 23.48% with a V OC of 1.18 V, a short-circuit current density (J SC ) of 25.1 mA/cm 2 and a fill factor (FF) of 78.4%. The integrated J SC obtained from the EQE measurement was 24.69 mA/cm 2 for the device, which is in accordance with the value determined from the J/ V scan (Fig. S2). Compared with the reference device PCE of 20.91% (Supplementary Table S1), the performance showed a 12.3% enhancement with suppressed hysteresis, which can be attributed to improved electron transport in the ETL and fewer electron accumulation at the interfaces of the ETL.
The structural properties of the SnO 2 ETL were first studied employing grazing incidence X-ray diffraction (GIXRD). At a low incident angle ω of 0.2°the signals related to the SnO 2 crystallites were observed with distinct diffraction peaks at 25.95°, 33.31°and 50.82°(Supplementary Figure S3A) whose positions are coherent with the tetragonal rutile structure of SnO 2 (ref. ICDD PDF 01-071-5,323). When increasing the incident angle ω from 0.2°to 0.4°, the ITO peaks became visible due to the larger depth probed, making it easy to distinguish them from the SnO 2 -related peaks. Compared with P-SnO 2 , films of G-SnO 2 showed no difference in the peak positions (Supplementary Figure S3B), confirming that Gua-Cl doping does not change the structural properties of the ETL. Next, ultraviolet photoelectron spectroscopy (UPS) was carried out to analyze the energy level positions of P-SnO 2 and G-SnO 2 . The work functions (Φ) of P-SnO 2 and G-SnO 2 were estimated from the secondary electron cutoff as −4.23 and −4.41 eV, respectively (Figure 2A). Plotting the UPS near the Fermi levels (E F ) allowed calculating the valence band maxima (VBM) with respect to E F (−3.73 and −3.69 eV for G/P-SnO 2 ) ( Figure 2B), resulting in of E VB = −8.14 and −7.92 eV vs. vacuum level, respectively (cf. Supplementary Figure S6). Notably, the Fermi level of the G-SnO 2 is situated deeper than that of the intrinsic P-SnO 2 layer, suggesting n-type doping and leading to a slight band bending at the interface with the perovskite layer. Figure 2C shows Tauc plots of each SnO 2 film on glass substrates. The optical bandgap values of G-SnO 2 and P-SnO 2 were deduced as 3.90 and 3.88 eV, respectively, i.e., essentially unchanged after doping. From these results, and combined with the conduction band minimum (CBM) of CsFAMA perovskites of −4.05 eV calculated by Abate and coworkers, (Wang et al., 2020b), an energy level diagram was constructed as shown in Figure 2D. As one important feature, an energy offset at the ETL/ perovskite interface of 0.19 eV is obtained with G-SnO 2 , while P-SnO 2 results in a flat alignment. Therefore, electron extraction is enhanced with G-SnO 2 because of the higher energetic driving force, , and a gradient-type band alignment between the perovskite layer and ITO is achieved.
The morphology of the perovskite absorber layer was studied by SEM. It can be seen from Figures 3A, B that the perovskite films on P-SnO 2 layers have crystallite sizes of around 200-400 nm, while much larger sizes of 500-800 nm were formed on G-SnO 2 layers, leading to the grain size distributions depicted in Supplementary Figure S7. As visible in the cross-section view, the perovskite films grown on both types of SnO 2 showed densely packed, highly crystalline grains ( Figure 3D). The differences observed between the two cases clearly demonstrate the direct influence of the underlying ETL layer, generating nucleation centers and promoting perovskite growth, on the crystalline quality of the perovskite film as has been shown in many literature examples (Bouchard et al., 2017;Dunlap-Shohl et al., 2019;Medjahed et al., 2020). The different morphology of the perovskite films having the same thickness ( Figure 3C, D) leads to a slight increase in the absorption of the layer grown on G-SnO 2 (Supplementary Figure  S8), which is beneficial for enhancing J sc . This behavior is in accordance with reported results comparing the light absorption properties of perovskite thin films constituted of different grain sizes. (Xiao et al., 2014).
To assess the surface morphology and roughness of the underlying ETL layer, atomic force microscopy (AFM) measurements were conducted. As shown in Figure 3F, the roughness of the G-SnO 2 film (RMS = 2.52 nm) is slightly larger than that (2.39 nm) of the P-SnO 2 film (cf. Supplementary Table S3), but the G-SnO 2 film shows a denser surface morphology with improved homogeneity. The upper perovskite layer exhibited a reduced RMS roughness (cf. Supplementary Table S4), from 23.6 nm grown on a P-SnO 2 film to 16.3 nm on G-SnO 2 . It can therefore be concluded that guanidine hydrochloride doped SnO 2 favors the formation of a dense and smooth perovskite layer which is beneficial for the subsequent deposition of the HTL. Moreover, the large crystallite size contributes to the suppression of trap states generated at grain boundaries.
Frontiers in Materials frontiersin.org X-ray diffraction patterns of the perovskite layers on the different ETLs ( Figure 4A) exhibit the main peaks located at 14.06°, 28.43°, and 31.84°, which are assigned to the (001), (002), and (012) reflections of the cubic crystalline phase of CsFAMA (space group m-3m). A strong diffraction peak at 12.68°originates from excess PbI 2 used in the perovskite preparation, which is known to passivate the grain boundaries. (Parida et al., 2021). Perovskite growth on P-SnO 2 resulted in a lattice constant of 6.278 (2) Å, while growth on G-SnO 2 led to a slightly decreased value of 6.274 (2) Å, as visible by the shift of the diffraction peaks to larger angles ( Figure 4B). Figure 4C depicts the evolution of the (00l) peak family for perovskites on P-SnO 2 and G-SnO 2 ETLs (cf.   Table S5). Due to the non-negligible contribution of the instrumental resolution to the broadening of the Bragg peaks, a quantitative analysis was not possible. However, considering that the crystallite size is >100 nm as determined from SEM analyses, the observed increased FWHM of the diffraction peaks denotes the presence of distortion (strain) in the lattice for both types of ETL. Importantly, the smaller FWHM increase observed for the perovskite grown on G-SnO 2 is a sign of reduced strain and improved crystalline quality compared to the perovskite layer on P-SnO 2 .
To further study the influence of the observed perovskite morphology and strain on the carrier recombination processes, we investigated the incident light intensity dependence of V OC ( Figure 4D). If the slope of V OC versus light intensity is greater than kT/q (q: electron charge, k: Boltzmann constant, T: Kelvin temperature), V OC is strongly dependent on the incident light intensity and additional recombination caused by strain is involved. Here, the P-SnO 2 devices exhibit a slope of 1.94 kT/q, while G-SnO 2 devices show a significantly smaller value (1.52 kT/q), indicating that trap-assisted recombination was effectively inhibited by reduced strain, which is consistent with the fitted recombination resistance from electrochemical impedance spectroscopy (Supplementary Figure S9). This assumption is further supported by steady-state and time-resolved photoluminescence (TRPL) spectroscopy, as shown in Figure 4E. In both cases, the perovskite film gives rise to a broad PL signal peaking at 762 nm, but the PL intensity of the film grown on the G-SnO 2 ETLs is significantly lower than that involving P-SnO 2 , which is a sign of enhanced electron injection. Analysis of the TRPL data ( Figure 4F) results in a shorter excited state lifetime for the ITO/G-SnO 2 / perovskite stack (τ = 318 ns) than for the sample comprising the P-SnO 2 ETL (τ = 415 ns), demonstrating the faster electron extraction from the perovskite film in the former case. Taken together, these data demonstrate that doping of the ETL with Gua-Cl leads to reduced charge accumulation and recombination losses at the perovskite/SnO 2 interface, which is highly beneficial for solar cell performance.
To assess the reproducibility of the PSCs, we tested 25 devices using each type of ETL ( Figure 5A). All solar cell parameters of the devices incorporating G-SnO 2 are improved, while the main contribution stems from enhanced J SC (23.4 vs. 24.9 mA/cm 2 ), originating from the concomitant reduction of the trap state density and the modulated band structure as discussed above. The statistical analysis yields an average and champion PCE of 22.86% and 23.48%, respectively, compared to 19.85% and 20.91% for the P-SnO 2 cells. These results demonstrate the high reproducibility and effectiveness of the employed strategy. Figure 5B exhibits the stabilized power output at the maximum power point (MPP) in a nitrogen-filled glovebox, and no change was visible after MPP tracking for 120 s for both types of PSC. The stabilized PCE of the G-SnO 2 -based cell is 23.2%, in agreement with the PCE obtained from the J/V curve. Finally, the stability of the non-encapsulated PSCs was studied under a dry air environment ( Figure 5C). The G-SnO 2 -based devices retained 96.1% of their initial PCE after 1,250 h, while the PCE of the P-SnO 2 based PSCs shrank to 89.9% under the same conditions. The observed stability enhancement is ascribed to the enhanced carrier extraction capacity of G-SnO 2 and the improved perovskite crystalline quality.

Conclusion
In summary, we presented a simple strategy for enhancing the performance of planar PSCs by doping the SnO 2 ETL with guanidinium chloride. The observed improvements are related to both the modification of the ETL intrinsic properties and the enhancement of the crystalline quality of the perovskite layer. More specifically, the modified G-SnO 2 ETL exhibited higher electron mobility and reduced defect state density than the pristine P-SnO 2 ETL, resulting in reduced interfacial charge carrier recombination. In addition, G-SnO 2 has a better-matched band alignment with the perovskite absorber further fostering efficient charge extraction. On the other hand, favored by the interaction between the guanidinium cation and halide anions from the perovskite precursors, the use of G-SnO 2 also induces the growth of a perovskite layer exhibiting significantly increased grain size (around 600 nm vs. 300 nm) and lower strain, while maintaining a dense packing and smooth surface. As a result of this synergistic behavior, the PSCs using G-SnO 2 exhibited an average PCE enhancement of 3% compared to devices with P-SnO 2 , reaching a maximum value of 23.48%. Combined with appealing stability and the possibility of low-temperature (80°C) processing, the proposed strategy also facilitates the realization of PSCs on flexible substrates and their integration into tandem devices.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
Benjamin Grévin is thanked for his valuable help with AFM analyses. The authors acknowledge the French Research Agency ANR for financial support (Grant PERSIL ANR-16-CE05-0019-02), and the LABEX Lanef (ANR-10-LABX-51-01) for its support with mutualized instrumentation. The PhD thesis of YLi was funded by UGA IDEX (project IRS C-Super), which is gratefully acknowledged.

Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.