Fluorine ion induced phase evolution of tin-based perovskite thin films: structure and properties

To study the effect of fluorine ions on the phase transformation of a tin-based perovskite, CsSnI3−x(F)x films were deposited by using thermal vacuum evaporation from a mixed powder of SnI2, SnF2 and CsI, followed by rapid vacuum annealing. The color evolution, structure, and properties of CsSnI3−xFx films aged in air were observed and analyzed. The results showed that the colors of the films changed from black to yellow, and finally presented as black again over time; the unstable B-γ-CsSnI3−xFx phase transformed into the Y-CsSnI3−xFx phase, which is then recombined into the Cs2SnI6−xFx phase with the generation of SnO2 in air. Fluorine dopant inhibited the oxidation process. The postponement of the phase transformation is due to the stronger bonds between F and Sn than that between I and Sn. The color changing process of the CsSnI3−xFx films slowed that the hole concentrations increased and the resistivities decreased with the increase of the F dopant ratio. With the addition of SnF2, light harvesting within the visible light region was significantly enhanced. Comparison of the optical and electrical properties of the fresh annealed CsSnI3−xFx films showed that the band gaps of the aged films widened, the hole concentrations kept the same order, the hole mobilities reduced and therefore, the resistivities increased. The double layer Cs2SnI6−xFx phase also showed ‘p’ type semi-conductor properties, which might be due to the incomplete transition of Sn2+ to Sn4+, i.e. Sn2+ provides holes as the acceptor.


Introduction
Within the past decade, inorganic-organic hybrid perovskite solar cells have drawn researchers' attention for their enhanced efficiencies due to their superior long diffusion length, with a conversion efficiency of approximately 3.8% in 2009 1 to greater than 23.32% today. 2 Since the early reports of lead (Pb) 3 and tin (Sn) 4 based perovskite solar cells, the efficiencies of Pb based perovskite solar cells have increased to nearly 22%, which exceeds those of poly-silicon solar cells. 5 Although most studies are focused on MAPbX 3 perovskite materials, tin-based perovskite has attracted more and more attention due to the low toxicity of tin. 6 So far, the literature 7,8 has reported that both Sn doped and the entirely Sn based perovskite solar cells (ASnX 3 ) are not as efficient as Pb-based perovskite solar cells due to two primary reasons: (1) tin is easily oxidized from Sn 2+ to Sn 4+ oxidation states whilst exposed to air. 9,10 The diffusion length of the photogenic carriers is limited because of too many p-type carriers produced by the self-doping effect within the tinbased perovskite materials; (2) SnI 2 and MAI (CH 3 NH 3 I À ) reacts quickly and rapidly crystallizes during spin-coating process 9,11 and therefore, it is difficult to control the crystallinity of perovskite, lead to poor coverage and uniformity of the lm. To combat this stable and defect free ASnX 3 lms are required.
Inorganic materials generally have higher stability than organic materials. For this reason, the studies on inorganic perovskite materials have been proposed. As the standard ASnX 3 perovskite, the schematic crystal structure of B-g-CsSnI 3 , Y-CsSnI 3 and Cs 2 SnI 6 are shown in Fig. 1, in which the B-g-CsSnI 3 phase is black with a three dimensional perovskite structure, 12 the Y-CsSnI 3 phase is yellow with a one-dimensional double-chain structure, 13 and Cs 2 SnI 6 is black with double layer perovskite structure. The unstable B-g-CsSnI 3 phase promptly transforms to the Y-CsSnI 3 phase, then forms into the Cs 2 SnI 6 phase in air, accompanied with the formation of SnO 2 . The compound of Cs 2 SnI 6 exhibits its stability in damp air due to the stable Sn 4+ state. The 'vacancy ordered' double layer perovskite structure is formed by the missing half of the Sn atoms located in the center of octahedral, and then reconstitutes to the discontinuous regular octahedral structure (SnI 6 ). [2][3][4][5][6][7][8][9][10][11][12][13][14] ASnX 3 perovskite structure (A ¼ metal or NH 4 and X ¼ halogen family) generally consists of (SnX 3 ) À and M + , in which Sn 2+ is unstable. Vilminot originally discussed evolution of the ionic-conductivity of the MSnF 3 phases (M ¼ Na, K, Rb, Cs, NH 4 , Tl) in 1985. 15 For example, CsSnF 3 exhibits a crystal structure consisting of isolated (SnF 3 ) À anionic polyhedral, 16 shown in Fig. 1d. Each Sn 2+ cation in CsSnF 3 is bonded to three uorine atoms in a distorted triangular pyramidal coordination environment, i.e. the lone pair of 5s electrons of tin(II) and three pairs of electrons, shared unequally with the uorine ions, occupied on the most stable conguration sp 3 hybridization orbits. 17 Researchers and experiments have shown that the tolerance factors of the stable perovskite materials should be between 0.78-1.05, 18 in which the tolerance factor, t ¼ (R A + R X )/O2(R B + R X ). In the formula, R A is the radius of monovalent cation, R B , the radius of divalent metal cations and R X , the radius of the halide ion. If there are multiple ions in the A and/or B positions, the average radius should be taken. The perovskite structure tends to be stable when t is close to 1. By calculation, the 't' of CsSnI 3Àx F x is within 0.8734-0.9587, and t gets larger as the F ratio increases. That means that the perovskite structure of CsSnF 3 is more stable than that of CsSnI 3 , and F doped in CsSnI 3 may delay the phase transformation of CsSnI 3 .
The crystal parameters of the CsSnI(F) 3 phases are shown in Table 1. The Sn-I bond length is stretched whilst the B-g-CsSnI 3 phase transformed into Y-CsSnI 3 , and then shortened along with the initial formation of the stable Cs 2 SnI 6 during the process of tin oxidization. There are many factors affecting bond length and bond energy, such as atomic radius, distance between nuclei, repulsive force between lone pair electrons, feedback bond and so on. In ionic compound, ion radius is the main contributor to bond length. Therefore, the Sn-F bond length in CsSnF 3 is less than the bond length of Sn-I in CsSnI 3 , the difference between them is 1.1321Å due to the much smaller ionic radii of F (1.33Å) compared to I (2.20Å). Generally, the order of stability of the halide complexes of tin(II) is F > Cl > Br > I, 17 tin(II) preferentially bonds with uoride ion. Hence, SnF 2 was introduced as the doping source to improve the stability of tin-based perovskite materials. Meanwhile, as an antioxidant, uoride ion inhibits the oxidation process of Sn 2+ , 10,19 the uorine also reduces the densities of the materials, which improves their performance when used as a photovoltaic material. 20 Even though the main method to prepare CsSnI 3 lms is still the one-step solution based process, it is difficult to produce dense, pinhole free lm due to the rapid crystallization of tinbased perovskite. 9,[20][21][22] The lms prepared by this method are very sensitive to lm formation conditions, such as annealing temperature, 23,24 solution concentration, 25,26 precursor solution composition 27,28 and solvent selection. [29][30][31] Due to the evaporation of the solvent and the volatilization of the materials during the process of annealing, the crystal tends to easily aggregate and shrink, and the morphology of the lms are mainly clusterlike and needle-like, so it is easy to cause the devices' efficiencies to be uneven. The thermal vacuum evaporation technique is an effective approach to prepare high coverage homogeneous thin lms and has been widely used in lead-based perovskites. 32,33 Therefore, tin-based perovskite lms should be possible to be prepared by this technique. 10,34 Here, the mixture powder of SnI 2 and CsI were evaporated onto the glass slide substrates by thermal vacuum evaporation method to form fully covered, dense, pinhole free CsSnI 3 perovskite lm. Also, by adding SnF 2 powder into the mixture, the CsSnI 3Àx F x lms were prepared as well. The evolution of the color, structure and properties of the annealed freshly and annealed aged CsSnI 3Àx F x lms over time was observed, measured and analyzed to explore the effect of F doping.
The process of sample preparation was shown in Fig. S1. † Weighted SnI 2 , SnF 2 and CsI powder was mixed in a mortar and placed in a tungsten boat of size 50 Â 15 Â 2 mm, refer to Table  2. The AC power (50 Hz frequency) was applied to the tungsten boat through two connected electrodes, the glass slides, CAT. no. 7101 with a size of 25.4 Â 76.2 mm, were ultrasonically cleaned in acetone for 900 s, diluted by deionized water and ethanol before being loaded into the DM 450C vacuum chamber. The glass slide was held above the tungsten boat at a separation of 150 mm. The chamber was pumped down to 2 Â 10 À3 Pa and the SnI 2 , SnF 2 and CsI was evaporated at a voltage of 70 V and current of 140 A. The evaporation process referred to Tong. 35 The lms were nally annealed at 210 C for 4 min in argon gas by the Rapid Thermal Processor of RTP-500V.

Measurement techniques
The thicknesses of the lms were measured by using a KLA-Tencor Alpha-step D-100 type prolometer on a step created on the lms by masking the glass substrates. The phases of the CsSnI 3Àx F x lms were measured by X'Pert powder X-ray diffractometer (XRD) in glancing angle scanning mode at 0.5 incident angle with Cu Ka X-ray from 10 to 80 , and analyzed by High-Score soware. 36 The electrical and optical properties of the CsSnI 3Àx F x lms were measured by a HALL 8800 Hall Effect Measurement device and a CARY 5000 UV-Vis-NIR spectrometer over the wavelength range of 300 to 1000 nm, respectively. AFM tests were performed in ambient conditions at room temperature with scanning probe microscopy (Bruker, Multimode 8 with controller V). A Pt/Ir-coated tip on a Si cantilever (tip radius of 20 nm, force constant of 2.8 Nm À1 and a resonant frequency of 75 kHz) was used to characterize the topography of the lms. The typical tip-scanning velocity was 2 mm s À1 .

Morphological structure
To demonstrate the macro-evolution process of the uorine doped cesium tin iodine (CsSnI 3Àx F x ) lms, photographs of the colors of the freshly annealed and aged CsSnI 3Àx F x lms exposed to ambient air were shown in Fig. 2a. With increasing  time, the colors of the lms changed from black to yellow, and nally presented as black again. These colors served as a good indication of the oxidation progress of B-g/Y-CsSnI 3 to Cs 2 SnI 6 in air. 37 It is evident from a comparison of the photographs in Fig. 2a that doping F slowed down the transformation process of the perovskite phases, since the color of F doped lms were predominantly yellow aer six hours exposure in air, whereas that of the CsSnI 3 lm without uorine dopant had almost completely blackened. The morphologies of the corresponding CsSnI 3Àx F x lms were examined by AFM, respectively (see Fig. 2b-e). Strikingly, no pinhole appeared in entire scope in the CsSnI 3Àx F x lms. Furthermore, the grain sizes became ner as the amount of doped uorine increased.

Phase structure
Corroborating evidence for the process of phase transition in CsSnI 3Àx F x lms \exposed in air was provided by X-ray diffraction (XRD), shown in Fig. 3a and b, which were consistent with the variations in morphology. Firstly, for the annealed CsSnI 3Àx F x lms (see Fig. 3a   (vaporization point 1553 K in ambient air) in the process of thermal evaporation. The variations of the XRD patterns show that uorine doping obstructed the transformation from B-g-CsSnI 3 to Y-CsSnI 3 perovskite phases. Aer aging seven days in air, the lms returned to a light black color. This phenomenon is consistent with the trend of Qiu's experiment. 37 The new diffraction peaks (see Fig. 3b) at 26.44 , 30.61 , and 43.91 were attributed to the (222), (004) and (044) orientations of Cs 2 SnI 6 phase (Ref. Code: 00-051-0466). However, the aged XRD patterns showed (120), (121), (130) and (210) planes of the Y-CsSnI 3 phase. Furthermore, several diffraction peaks at 27.57 , 39.41 , and 26.58 assigned to CsI and SnO 2 (Ref.Code: 00-077-0448) were observed as well. Again, the Y-CsSnI 3 phase was much obvious with the increase of F dopant ratios, which meant F doping delayed the phase transformation, and Y phase existed in the highly F doped lms even aer aging for seven days.
Based on these ndings via the above XRD data, partial CsI, SnI 2 , and SnO 2 peaks were observed during the evolution process of the CsSnI 3Àx F x lms, the chemical reactions involved may be as follows: Therefore, it is clear that this process was due to oxidation of Sn in the compounds. 6 Sn 2+ ions are quite sensitive to external oxygen, especially in humid environments, which can be oxidized to more stable Sn 4+ analogues. This oxidation process may fundamentally destroy the charge neutrality of the CsSnI 3 perovskite structure and lead to phase transition. The Sn-F bonds form when uorine replaces iodide in the lattice. The Sn-F bond length is shorter than that of Sn-I, resulting in a more stable crystal structure and weakening the role of oxygen, refer to Fig. S2. † In short, doping with uorine slowed down the oxidation progress of the Sn 2+ in CsSnI 3 into Sn 4+ . The B-g-CsSnI 3Àx (F) x phase eventually transferred into Cs 2 SnI 6Àx (F) x and was accompanied by the formation of SnO 2 over time, but the formation processes are delayed by the additional F dopant. The remnant Y-CsSnI 3 phase proved not only to delay the phase transition, but also the existence of Sn 2+ . The hybrid of Sn 4+ with Sn 2+ resulted in the Cs 2 SnI 6Àx (F) x lms showing their p type property.

Electrical properties of CsSnI 3Àx F x lms
The electrical properties and thickness of the CsSnI 3Àx F x lms at room temperature were measured by a HALL 8800 Hall Effect Measurement device and KLA-Tencor Alpha-step D-100 type prolometer respectively. The results were shown in Table 3 for the freshly annealed and Table 4 for those aged seven days aer annealing. In this case, the thicknesses of the lms were used to calculate the resistivity of the lms. Both the annealed and aged CsSnI 3Àx F x lms were p type semiconductors with the holes provided by the Sn vacancies. Intrinsic defects such as Sn vacancies in the ternary Cs-Sn-I system gave rise to p-type conductivity 20 and DFT calculations have shown that the formation energy of V Sn defects was the lowest among all defects. 38 In Table 3, the freshly annealed CsSnI 3 lm exhibited the carrier densities of $10 14 cm À3 , and the carrier density of CsSnI 3Àx F x lms increased to $10 16 cm À3 with the increase of SnF 2 content. It's also worth noting that the resistivities of CsSnI 3Àx F x lms were 1-2 orders of magnitude smaller than that of undoped lm. The reduction of the Sn vacancy concentrations can be attributed to the strong bonding energy between F and Sn.
Aer aging in air for seven days, the electrical properties of the CsSnI 3Àx F x lms kept the same order. As described in Section 3.2 Phase structure, the main composition of the CsSnI 3Àx F x lms was Cs 2 SnI 6 . This is accompanied by the generation of a mass of Sn vacancies. The carrier densities of the aged CsSnI 3Àx F x lms were almost at the same level in comparison to those of the annealed CsSnI 3Àx F x lms. The resistivities of CsSnI 3Àx F x lms were still relatively small. Aer aging for seven days, the phase transformation occurred, the oxidation of tin was completed, and the double layer 'vacancy ordered' phase Cs 2 SnI 6Àx F x formed. As the results, the Sn vacancies of the aged lms were high, and their carrier density increased sharply. However, the rate of increase of the charge carriers of the aged lms decreased with the increase of SnF 2 content. The phases were stable due to F doping, the creation processes of the Sn vacancies were delayed and therefore, the changes of the electrical properties were slower. This fully illustrates SnF 2 hindering the oxidation process of Sn 2+ , which blocked the formation of Cs 2 SnI 6 .
Since the carrier (hole or electron) concentration of semiconductor depends on the inherent defect concentration, the control of carrier concentration is a necessary condition to optimize the performance of solar cells. It works better as the hole-transport material (HTM) when hole concentration and conductivity are high, but it may work better as the light absorber material (LAM) when hole concentration and conductivity are moderate. 39 From the data of electrical properties, by tuning the F doping amount, hole concentration and

Optical properties
The absorptive spectra of the annealed and aged CsSnI 3Àx F x lms within the range of 400-1400 nm were presented in Fig. 4a and c, respectively. Within the UV range, the absorptivity of the lms was almost the same. The absorption edges of the annealed and aged CsSnI 3Àx F x lms exhibited obvious differences. With the addition of SnF 2 , the light harvesting in the visible light region of the solar spectrum of the CsSnI 3Àx F x lms was signicantly enhanced. These differences indicate the difference of the band gap between samples. The spectra of the absorption via optical band gaps of the annealed and aged lms were shown in Fig. 4b and d, which were calculated by the formula of ahq ¼ A(hq À E g ) 1/2 . The Tauc plot 40 is used to evaluate the optical energy gap. The bandgaps of all the CsSnI 3Àx F x lms are between 1.25 eV and 1.3 eV as shown in Fig. 4b which was consistent with those previously reported. 41 The optical spectra of CsSnI 3Àx F x lms under exposure to ambient air for seven days are shown in Fig. 4c and d. The absorption edges of all the CsSnI 3Àx F x lms exhibited blue shi. All the bandgaps for the F doped lms got smaller, and was close to 1.48 eV derived in Qiu's report. 37 This is a result of SnF 2 eliminating the formation of weak unidentied reections due to the Y-CsSnI 3 phase. 42 Aer the CsSnI 3Àx F x lms under exposure to ambient air for seven days, the Y-CsSnI 3 phase only was shown in the highly F doped lms, which the fully illustrates the role of SnF 2 . The transmittance of the Cs 2 SnI(F) 6 lms within visible wavelength were strongly affected by doping F ions due to the decomposition of CsSnI 3Àx F x . It was further shown that SnF 2 blocked the formation of Cs 2 SnI 6 by the changes of bandgap. Note that the sunlight absorption in the red and near-infrared regions has been a challenge for solar cells, and the excellent optical properties of CsSnI 3Àx F x lms made up for this defect.

Conclusion
CsSnI 3 transformed into stable double layer perovskite Cs 2 SnI 6 phase in air was investigated for potential solar cell applications. By conducting a series of experiments, it was determined that the phase transformation in air can be slowed down by doping F ions into CsSnI 3 . The mechanism of phase transition delay is that Sn 2+ preferentially bonds with F À , and thus the process of Sn 2+ in CsSnI 3 to be oxidized into Sn 4+ was slowed down. The B-g-CsSnI(F) 3 phase eventually formed Cs 2 SnI(F) 6 accompanied by the formation of the SnO 2 phase over time, but the transition processes are delayed by the additional F dopant. Furthermore, SnF 2 improved the carrier concentration and conductivity of the lm, in which the resistivities of CsSnI 3Àx F x lms were less than 10 U cm except CsSnI 2.67 F 0.33 . The CsSnI 3Àx F x lms also absorbed more visible and infrared light with increased F doping than the pure CsSnI 3 lm. It was further shown that SnF 2 blocked the formation of Cs 2 SnI 6 by comparing the changes of bandgaps. The usage of CsSnI 3Àx F x lms as HTM or LAM in solar cells is envisioned to be a promising method of improving efficiency.

Conflicts of interest
There are no conicts to declare.