Modulating Surface Morphology Related to Crystallization Speed of Perovskite Grain and Semiconductor Properties of Optical Absorber Layer under Controlled Doping of Potassium Ions for Solar Cells

Perovskite thin films with excellent optical semiconductor and crystallization properties and superior surface morphology are normally considered to be vital to perovskite solar cells (PSCs). In this paper, we systematically survey the process of modulating surface morphology and optical semiconductor and crystallization properties of methylammonium lead iodide film by controlling doping of K+ for PSC prepared in air and propose the mechanism of large K+-doped perovskite grain formation related to crystallization speed. The increase in the crystallization speed leads to the production of large grains without localized-solvent-vapor (LSV) pores via moderate doping of K+, and the exorbitant crystallization speed induces super large grains with LSV pores via excessive doping of K+. Furthermore, the semiconductor properties (absorption band edge wavelength, PL emission peak wavelength, energy band gap) of perovskite film can be significantly tuned by controlled doping of K+. The investigation of the detailed process of modulating surface morphology and semiconductor properties of perovskite thin film by controlled doping of K+ may provide guidance and pave the way for superior component design of absorption materials for cost-efficient PSCs.


Results and Discussion
The K + doping contents of 1.  Figure 1, where we can see clearly the process of modulating surface morphology of the perovskite thin film via controlled doping of K + . In Figure 1e-h, we can see clearly the perovskite layer, M-/C-TiO 2 (Mesoporous-TiO 2 and Compact-TiO 2 ) layer and FTO layer from the top to the bottom. Although the surface of MAPbI 3 +0M (presented in Figure 1a) is relatively flat, there are many pores (presented in yellow circle, Figure 1e) in the interface between MAPbI 3 +0M and M-/C-TiO 2 . In addition, the sizes of grains are relatively smaller and there are more vertical grain boundaries in MAPbI 3 +0M (presented in the white circle, Figure 1e). Relative to MAPbI 3 +0M, the surface of MAPbI 3 +0.6M (presented in Figure 1b) is rough, but pores are not present in the interface between perovskite layer and M-/C-TiO 2 , which means the doping of K + can passivate the interface between MAPbI 3 +0M and M-/C-TiO 2 . Furthermore, the sizes of grains are bigger, and there are fewer transverse grain boundaries (presented in Figure 1b). As for MAPbI 3 +0.9M (Figure 1c,g), the surface has better flatness and grains are larger than for MAPbI 3 +0.6M, so that the transverse grain boundaries are greatly reduced and vertical grain boundaries are almost gone, which lead to a great improvement in its efficiency. From Figure 1d, we can find a super large grain (>4 µm in the longest direction), but there are LSV pores (presented in blue circle), boundary gaps (presented in yellow circle) and unidentified square protrusions (presented in red circle) on the surface of MAPbI 3 +1.2M. It is more important to note that there are also LSV pores (presented in blue circle) inside MAPbI 3 +1.2M according to Figure 1h, which greatly reduces the efficiency of the device. These LSV pores are caused by the inefficient discharge of the solvent from the interior of crystallizing perovskite due to the exorbitant perovskite CS. Through the comparison of these cross-sectional SEM images (Figure 1e-h), we can find that MAPbI 3 +1.2M becomes very thick (~900 nm). Furthermore, the perovskite grain sizes increases (~500 nm,~700 nm,~2 µm and~4 µm, on average, corresponding MAPbI 3 +0M, MAPbI 3 +0.6M, MAPbI 3 +0.9M and MAPbI 3 +1.2M, respectively) with the increase in the doping concentration of K + according to the comparison of these surface SEM images (Figure 1a-d).
The perovskite grain sizes increase with the increase in the doping concentration of K + so that the number of grain boundaries of perovskite films decreases. In addition, the decomposition of perovskite films begins at the boundaries of perovskite grains, so the stability of devices increases with the increase in the doping concentration of K + . perovskite grain sizes increase with the increase in the doping concentration of K + so that the number of grain boundaries of perovskite films decreases. In addition, the decomposition of perovskite films begins at the boundaries of perovskite grains, so the stability of devices increases with the increase in the doping concentration of K + . According to these experimental results, we propose the following mechanism (presented in Figure 2) of large K + -doped perovskite grain formation related to CS. In Figure 2, the red arrow represents the solvent evaporation (SE) direction and the small black spot represents seed crystal, which is the root of the formation of perovskite grain. We assume that the distributions of seed crystals in all samples are the same during the initial stage. Lower CS (corresponding to the absence of doping) leads to relatively more spaces to form seed crystals in the perovskite solution due to smaller crystallizing grains during the intermediate stage, which results in smaller grains and more grain boundaries during the final stage eventually. Suitable CS causes larger crystallizing perovskite grains and less spaces for the formations of new seed crystals during the intermediate stage via moderate doping of K + , and finally, the large grains without LSV pores formed. However, excessive doping can cause exorbitant CS and leads to super large crystallizing grain during the intermediate stage, so that there are almost no spaces to form new seed crystals and the solvent can only be evaporated upwards for filling of the whole space by the transverse crystallizing grains, which cause the solvent to not effectively discharge from the interior of crystallizing perovskite; as a result, LSV pores are formed inside the perovskite thin films and on its surface during the final stage. The crystallization already starts when the seed crystal form and it may be during solvent evaporation when the spin-coating or the process of chlorobenzene addition or thermal annealing occur. Irrespective, the schematic mechanism in Figure 2 is adaptive. In addition, it is worth emphasizing that new seed crystals may form at all times, and once the seed crystals form, the process described by the schematic mechanism takes place.  According to these experimental results, we propose the following mechanism (presented in Figure 2) of large K + -doped perovskite grain formation related to CS. In Figure 2, the red arrow represents the solvent evaporation (SE) direction and the small black spot represents seed crystal, which is the root of the formation of perovskite grain. We assume that the distributions of seed crystals in all samples are the same during the initial stage. Lower CS (corresponding to the absence of doping) leads to relatively more spaces to form seed crystals in the perovskite solution due to smaller crystallizing grains during the intermediate stage, which results in smaller grains and more grain boundaries during the final stage eventually. Suitable CS causes larger crystallizing perovskite grains and less spaces for the formations of new seed crystals during the intermediate stage via moderate doping of K + , and finally, the large grains without LSV pores formed. However, excessive doping can cause exorbitant CS and leads to super large crystallizing grain during the intermediate stage, so that there are almost no spaces to form new seed crystals and the solvent can only be evaporated upwards for filling of the whole space by the transverse crystallizing grains, which cause the solvent to not effectively discharge from the interior of crystallizing perovskite; as a result, LSV pores are formed inside the perovskite thin films and on its surface during the final stage. The crystallization already starts when the seed crystal form and it may be during solvent evaporation when the spin-coating or the process of chlorobenzene addition or thermal annealing occur. Irrespective, the schematic mechanism in Figure 2 is adaptive. In addition, it is worth emphasizing that new seed crystals may form at all times, and once the seed crystals form, the process described by the schematic mechanism takes place. perovskite grain sizes increase with the increase in the doping concentration of K + so that the number of grain boundaries of perovskite films decreases. In addition, the decomposition of perovskite films begins at the boundaries of perovskite grains, so the stability of devices increases with the increase in the doping concentration of K + . According to these experimental results, we propose the following mechanism (presented in Figure 2) of large K + -doped perovskite grain formation related to CS. In Figure 2, the red arrow represents the solvent evaporation (SE) direction and the small black spot represents seed crystal, which is the root of the formation of perovskite grain. We assume that the distributions of seed crystals in all samples are the same during the initial stage. Lower CS (corresponding to the absence of doping) leads to relatively more spaces to form seed crystals in the perovskite solution due to smaller crystallizing grains during the intermediate stage, which results in smaller grains and more grain boundaries during the final stage eventually. Suitable CS causes larger crystallizing perovskite grains and less spaces for the formations of new seed crystals during the intermediate stage via moderate doping of K + , and finally, the large grains without LSV pores formed. However, excessive doping can cause exorbitant CS and leads to super large crystallizing grain during the intermediate stage, so that there are almost no spaces to form new seed crystals and the solvent can only be evaporated upwards for filling of the whole space by the transverse crystallizing grains, which cause the solvent to not effectively discharge from the interior of crystallizing perovskite; as a result, LSV pores are formed inside the perovskite thin films and on its surface during the final stage. The crystallization already starts when the seed crystal form and it may be during solvent evaporation when the spin-coating or the process of chlorobenzene addition or thermal annealing occur. Irrespective, the schematic mechanism in Figure 2 is adaptive. In addition, it is worth emphasizing that new seed crystals may form at all times, and once the seed crystals form, the process described by the schematic mechanism takes place.  To back up this theory that the CS of perovskite increases with the increase in the K doping amount, we analyze the results of previous research. Uz Zaman et al. prepared K-doped perovskite thin films on FTO substrates by spin coating [35], which are hydrophobic substrates, and the perovskite solution films must shrink after spin coating, unavoidably so that the final perovskite films cannot completely cover the substrates. We can find that the coverage area of perovskite thin films increases with the increase of the K doping amount according to their SEM results [35]. The coverage rate of perovskite thin film on substrate is mainly determined by the strength of the solution thin film shrinkage and the CS of perovskite, illustrated in Figure 3. In Figure 3, the red arrow marks the shrinkage direction of the perovskite solution thin film and the small black dot represents the perovskite seed crystal. In the process of perovskite solution film shrinking, perovskite crystals are also growing in areas with perovskite precursor solution, so that stronger shrinkage strength leads to a smaller coverage rate and faster crystallization speed leads to a greater coverage rate. The shrinkage strength of the perovskite solution film is mainly determined by the solvent and substrate; however, the solvents of the perovskite precursor solutions with different K content and substrates are exactly the same, so the shrinkage strengths of perovskite solution films are almost the same. Thus, the conclusion can be reached that different crystallization speeds lead to different perovskite film coverage rates, and the crystallization speed of perovskite increases with the increase in the K doping amount. The crystallization speed of perovskite we expect can make the grain grow to a large state without localized-solvent-vapor (LSV) pores or with small number of LSV pores of which the inducing efficiency attenuation is less than the benefit from the increase in grain size. According to SEM diagrams and JV curves, the expected speed is between the speed corresponding 0.9 M and the speed corresponding 1.2 M because MAPbI 3 +0.9M does not have LSV pores and MAPbI 3 +0.9M has LSV pores, which induce efficiency attenuation. Exorbitant crystallization speed can make the grain grow to a large state with more LSV pores of which the inducing efficiency attenuation is higher than the benefit from the increase in grain size, so the speed corresponding to 1.2 M is an exorbitant crystallization speed. To back up this theory that the CS of perovskite increases with the increase in the K doping amount, we analyze the results of previous research. Uz Zaman et al. prepared K-doped perovskite thin films on FTO substrates by spin coating [35], which are hydrophobic substrates, and the perovskite solution films must shrink after spin coating, unavoidably so that the final perovskite films cannot completely cover the substrates. We can find that the coverage area of perovskite thin films increases with the increase of the K doping amount according to their SEM results [35]. The coverage rate of perovskite thin film on substrate is mainly determined by the strength of the solution thin film shrinkage and the CS of perovskite, illustrated in Figure 3. In Figure 3, the red arrow marks the shrinkage direction of the perovskite solution thin film and the small black dot represents the perovskite seed crystal. In the process of perovskite solution film shrinking, perovskite crystals are also growing in areas with perovskite precursor solution, so that stronger shrinkage strength leads to a smaller coverage rate and faster crystallization speed leads to a greater coverage rate. The shrinkage strength of the perovskite solution film is mainly determined by the solvent and substrate; however, the solvents of the perovskite precursor solutions with different K content and substrates are exactly the same, so the shrinkage strengths of perovskite solution films are almost the same. Thus, the conclusion can be reached that different crystallization speeds lead to different perovskite film coverage rates, and the crystallization speed of perovskite increases with the increase in the K doping amount. The crystallization speed of perovskite we expect can make the grain grow to a large state without localized-solvent-vapor (LSV) pores or with small number of LSV pores of which the inducing efficiency attenuation is less than the benefit from the increase in grain size. According to SEM diagrams and JV curves, the expected speed is between the speed corresponding 0.9 M and the speed corresponding 1.2 M because MAPbI3+0.9M does not have LSV pores and MAPbI3+0.9M has LSV pores, which induce efficiency attenuation. Exorbitant crystallization speed can make the grain grow to a large state with more LSV pores of which the inducing efficiency attenuation is higher than the benefit from the increase in grain size, so the speed corresponding to 1.2 M is an exorbitant crystallization speed.  Figure 4a shows XRD pattern of perovskite thin films with different doping concentration of K + and we can find that the peak position of perovskite crystal has not been transferred, which means there is no change in the type of perovskite crystal structure via different concentration doping of K + [36]. In addition, the perovskite film exhibits pure tetragonal phase according to Figure 4a, which indicates that the K + enters into the perovskite lattice successfully [20]. From Figure 4b, we noticed that the perovskite crystallization is improved with the increase in the K doping amount, which indicates a longer carrier lifetime [20]. Compared to MAPbI3+0M, almost all other (040) peaks shift to lower degree in Figure 4c, which means K + mainly occupied interstitial sites in lattices [20].  Figure 4a shows XRD pattern of perovskite thin films with different doping concentration of K + and we can find that the peak position of perovskite crystal has not been transferred, which means there is no change in the type of perovskite crystal structure via different concentration doping of K + [36]. In addition, the perovskite film exhibits pure tetragonal phase according to Figure 4a, which indicates that the K + enters into the perovskite lattice successfully [20]. From Figure 4b, we noticed that the perovskite crystallization is improved with the increase in the K doping amount, which indicates a longer carrier lifetime [20]. Compared to MAPbI 3 +0M, almost all other (040) peaks shift to lower degree in Figure 4c, which means K + mainly occupied interstitial sites in lattices [20].  The Energy dispersive X-ray Spectroscopy (EDX) of K-doped perovskite crystallite films and the specific values of quantifying doping levels are presented in Figure 5 and Table 1, respectively. In the EDXs (Figure 5a-d), we can see that the peaks of K elements (presented in blue circle) increase with the increase in the doping concentration, which is consistent with the values in Table  1. The excess iodine coming from the KI solution can also affect the perovskite film performance, which is carried out by Equation (1) [37] and Equation (2) [37,38].
However, I3 − can effectively decrease the concentration of defects in organic-inorganic lead halide perovskite films [38].  The Energy dispersive X-ray Spectroscopy (EDX) of K-doped perovskite crystallite films and the specific values of quantifying doping levels are presented in Figure 5 and Table 1, respectively. In the EDXs (Figure 5a-d), we can see that the peaks of K elements (presented in blue circle) increase with the increase in the doping concentration, which is consistent with the values in Table 1. The excess iodine coming from the KI solution can also affect the perovskite film performance, which is carried out by Equation (1) [37] and Equation (2) [37,38].
However, I 3 − can effectively decrease the concentration of defects in organic-inorganic lead halide perovskite films [38].  The Energy dispersive X-ray Spectroscopy (EDX) of K-doped perovskite crystallite films and the specific values of quantifying doping levels are presented in Figure 5 and Table 1, respectively. In the EDXs (Figure 5a-d), we can see that the peaks of K elements (presented in blue circle) increase with the increase in the doping concentration, which is consistent with the values in Table  1. The excess iodine coming from the KI solution can also affect the perovskite film performance, which is carried out by Equation (1) [37] and Equation (2) [37,38].
However, I3 − can effectively decrease the concentration of defects in organic-inorganic lead halide perovskite films [38].    Figure 6a,b show the absorbance spectra and PL spectra of perovskite thin films with 0 M, 0.6 M, 0.9 M and 1.2 M, respectively. There are blue shifts of the absorption band edge wavelength and PL emission peak wavelength of MAPbI 3 +0.6M and MAPbI 3 +0.9M, and the red shift the absorption band edge wavelength and PL emission peak wavelength of MAPbI 3 +1.2M has taken place, compared to reference MAPbI 3 +0M, indicating that the energy band gap increases first and then decreases with the increase in the doping concentration of K + [5].   Figure 6a,b show the absorbance spectra and PL spectra of perovskite thin films with 0 M, 0.6 M, 0.9 M and 1.2 M, respectively. There are blue shifts of the absorption band edge wavelength and PL emission peak wavelength of MAPbI3+0.6M and MAPbI3+0.9M, and the red shift the absorption band edge wavelength and PL emission peak wavelength of MAPbI3+1.2M has taken place, compared to reference MAPbI3+0M, indicating that the energy band gap increases first and then decreases with the increase in the doping concentration of K + [5]. Schematic illustration for the process of modulating surface morphology related to CS and optical semiconductor properties of perovskite thin films via controlled doping of K + is shown in Figure 7. It gives a very detailed description of the process of modulating surface morphology (including grain size, surface flatness, transverse and vertical grain boundary quantity, internal LSV pores, surface LSV pores, pores in the interface between perovskite layer and M-/C-TiO2, thickness of perovskite thin film, boundary gap) and optical semiconductor properties (absorption band edge wavelength, PL emission peak wavelength, energy band gap) of perovskite thin film via controlled doping of K + prepared in air. It is worth emphasizing that the mechanism of large K + -doped perovskite grain formation is related to CS. The doping of K + leads to an increase in CS of perovskite grains during the intermediate stage, and results the formation of large perovskite grains during the final stage. However, excessive doping of K leads to a rapid perovskite crystallization speed that induces super large crystallizing grains and causes the solvent to not effectively discharge from crystallizing perovskite during the intermediate stage; as a result, LSV pores are formed inside the perovskite thin films and on its surface during the final stage.
The structure of the device is shown in Figure 8a. Figure 8b shows the J-V curves from RS for the devices with different doping concentrations of K + , measured under simulated sunlight (AM 1.5 G), and the detail photovoltaic parameters of PSCs is shown in the Table 2. Voc has increased significantly via controlling doping of K + , as a result of elimination of pores in the interface between the perovskite layer and M-/C-TiO2 layer. Due to larger grains in the MAPbI3+0.9M and MAPbI3+1.2M, the corresponding devices obtain higher Jsc. Although the absorption spectrum of MAPbI3+0.9M is relatively low compared to other perovskite films, its device yielded the best performance (especially the Voc and Jsc) due to prolonged carrier lifetime, improved surface Schematic illustration for the process of modulating surface morphology related to CS and optical semiconductor properties of perovskite thin films via controlled doping of K + is shown in Figure 7. It gives a very detailed description of the process of modulating surface morphology (including grain size, surface flatness, transverse and vertical grain boundary quantity, internal LSV pores, surface LSV pores, pores in the interface between perovskite layer and M-/C-TiO 2 , thickness of perovskite thin film, boundary gap) and optical semiconductor properties (absorption band edge wavelength, PL emission peak wavelength, energy band gap) of perovskite thin film via controlled doping of K + prepared in air. It is worth emphasizing that the mechanism of large K + -doped perovskite grain formation is related to CS. The doping of K + leads to an increase in CS of perovskite grains during the intermediate stage, and results the formation of large perovskite grains during the final stage. However, excessive doping of K leads to a rapid perovskite crystallization speed that induces super large crystallizing grains and causes the solvent to not effectively discharge from crystallizing perovskite during the intermediate stage; as a result, LSV pores are formed inside the perovskite thin films and on its surface during the final stage.
The structure of the device is shown in Figure 8a. Figure 8b shows the J-V curves from RS for the devices with different doping concentrations of K + , measured under simulated sunlight (AM 1.5 G), and the detail photovoltaic parameters of PSCs is shown in the Table 2. V oc has increased significantly via controlling doping of K + , as a result of elimination of pores in the interface between the perovskite layer and M-/C-TiO 2 layer. Due to larger grains in the MAPbI 3 +0.9M and MAPbI 3 +1.2M, the corresponding devices obtain higher J sc . Although the absorption spectrum of MAPbI 3 +0.9M is relatively low compared to other perovskite films, its device yielded the best performance (especially the V oc and J sc ) due to prolonged carrier lifetime, improved surface morphology and crystallinity, which means the photovoltaic properties of the devices are improved via appropriate doping of K + . A promising efficiency of 8.14% was achieved for the 0.9 M K + -doped device with the carbon/FTO CE, and these simple and low-cost preparation techniques of high-class perovskite thin films and carbon/FTO CE under ambient conditions are beneficial for promoting commercialization.

Conclusions
The LSV-pore-free perovskite grains became significantly larger via moderate doping of K + . The mechanism of large K + -doped perovskite grain formation is related to CS and the theory that the CS of perovskite increases with the increase of K doping amount was suggested and preliminarily confirmed. The detailed description of the process of modulating surface morphology (including grain size, surface flatness, transverse and vertical grain boundary quantity, internal LSV pores, morphology and crystallinity, which means the photovoltaic properties of the devices are improved via appropriate doping of K + . A promising efficiency of 8.14% was achieved for the 0.9 M K + -doped device with the carbon/FTO CE, and these simple and low-cost preparation techniques of high-class perovskite thin films and carbon/FTO CE under ambient conditions are beneficial for promoting commercialization.

Conclusions
The LSV-pore-free perovskite grains became significantly larger via moderate doping of K + . The mechanism of large K + -doped perovskite grain formation is related to CS and the theory that the CS of perovskite increases with the increase of K doping amount was suggested and preliminarily confirmed. The detailed description of the process of modulating surface morphology (including grain size, surface flatness, transverse and vertical grain boundary quantity, internal LSV pores,

Conclusions
The LSV-pore-free perovskite grains became significantly larger via moderate doping of K + . The mechanism of large K + -doped perovskite grain formation is related to CS and the theory that the CS of perovskite increases with the increase of K doping amount was suggested and preliminarily confirmed. The detailed description of the process of modulating surface morphology (including grain size, surface flatness, transverse and vertical grain boundary quantity, internal LSV pores, surface LSV pores, pores in the interface between MAPbI 3 +0M and M-/C-TiO 2 , thickness of perovskite thin film, boundary gap) and optical semiconductor properties (absorption band edge wavelength, PL emission peak wavelength, energy band gap) of perovskite thin film via controlled doping of K + prepared in air was presented. A promising efficiency of 8.14% in a size of 0.2 cm 2 was achieved for the device with inexpensive carbon/FTO CE under one sun illumination. Obviously, mass production of cost-efficient K + -doped PSCs, with spongy carbon/FTO CEs, under ambient atmosphere, is possible.

Conflicts of Interest:
The authors declare no conflict of interest.