Environment-friendly High Eciency CH3NH3PbI3 Perovskite Solar Cells Fabricated by Green Antisolvent Method

Green antisolvents propyl acetate and isooctane can promote the nucleation of PbI 2 particles and provide the heterogeneous nucleation site of perovskite crystal, so as to promote the rapid growth of perovskite crystal. The lms treated with propyl acetate and propylene glycol methyl ether have larger grains and lower root-mean-square value (RMS) than those treated with chlorobenzene. The full coverage perovskite lms with uniform grain size which is closer to the exciton diffusion length of perovskite can be obtained. Compared with the CH 3 NH 3 PbI 3 perovskite solar cells treated with chlorobenzene (power conversion eciency (PCE) is 17.86%), the best eciency of the devices treated with propylene glycol methyl ether is 21.60%, which is improved by 21%, which has certain reference value and guiding signicance for researchers to obtain environment-friendly high-quality perovskite solar cells in the future.


Introduction
Halogenated perovskite materials have excellent photoelectric properties, such as high extinction coe cient, long carrier diffusion length, high carrier mobility, high defect tolerance and controllable band gap, etc., which have attracted more and more attention of researchers all over the world. In the past ten years, the power conversion e ciency of the perovskite solar cells can be obtined from 3.8-25.5% [1][2][3][4][5] . Perovskite solar cells (PSCs) can be prepared by solution method, but it is inevitable that solution treated polycrystalline lms tend to have high-density defect states and show a large degree of electronic disorder. It leads to carrier recombination and increase of interface resistance. Defects with transition energy levels in the band gap will attract more energy electrons / holes as non radiative recombination centers. The Fermi level is pinned near the defect state. Non radiative recombination is induced by charge recombination of these defect states, which constitutes a main energy loss mechanism. They will mean reducing the lling factor (FF) and open circuit voltage (Voc) of the device. Theoretical modeling and experimental studies have shown that Voc in photovoltaic devices is divided by the quasi Fermi energy levels of electrons and holes in the whole, and the disorder of band tail defects or trap states caused by energy extends to the forbidden band, introducing additional electronic states. Nonequilibrium photogenerated electrons tend to relax downward. For these trap states, this will lead to lower splitting of the quasi Fermi level and smaller open circuit voltage. Due to the insertion of MAI into PbI 2 lattice, the penetration of methylammonium halide becomes a serious problem, which will lead to incomplete conversion. Due to the formation of perovskite structure dense layer of PbI 2 , MAI diffusion and insertion into deeper PbI 2 layer are prevented. In addition, residual and uncontrolled PbI 2 have adverse effects on carrier separation and transport and subsequent equipment performance carrier, and reproducibility and stability cannot be guaranteed. In order to solve this problem, Seok's team used PbI 2 (DMSO) complex to increase the conversion process, leading to high-performance solar cells. Huang and his colleagues developed a thermally driven cell, using the interdiffusion method to generate a compact MAPbI 3 layer.
Zhao and his colleagues added some Mai to PbI 2 precursors to produce adjustable volume expansion before Mai contact, so as to produce e cient and renewable solar cells. Improving the lm quality of perovskite lm (such as density, roughness and grain size) can effectively improve the energy conversion e ciency of the device. Researchers have developed a series of methods to prepare high-quality perovskite lms, such as one-step lm forming method [6][7][8][9][10] , two-step lm forming method [11][12][13][14][15] .
Among them, antisolvent method or antisolvent assisted crystallization (ASAC) is the most widely used method because of its low cost, simple process and good repeatability. [16][17][18] However, the antisolvent assisted crystallization method usually uses toxic solvents such as chlorobenzene, toluene, dichloromethane, etc. In general, researchers need to use glove boxes, fume hoods, gas masks and so on to isolate the damage from toxic solvents. If this method is applied to the large-area preparation of perovskite lms or the production line, it will inevitably increase the possibility of exposing the experimenters to toxic atmosphere. Therefore, looking for new green solvents that can replace toxic solvents without (or as little as possible) environmental side effects has become the focus of attention.
Green solvent generally refers to the unstable chemical properties of the solvent, which can be degraded by soil organisms or other substances, with short half-life and easy to decay into low toxic and non-toxic substances. In recent years, researchers have invested a lot of research work in the development of green solvent or solvent-free preparation process of perovskite lms, and made a lot of meaningful research results. [19] For example, Wang Shimin studied the effects of different antisolvents( toluene, chlorobenzene, ethyl acetate, diethyl ether and isopropanol) on the quality of perovskite lms and the performance of solar cells. At last, they concluded that ethyl acetate with low toxicity was the most suitable antisolvent. [20] Qin Yong systematically studied the in uence of green solvent anisole and toxic solvent chlorobenzene on the absorbance and crystallinity of perovskite lms. It was found that the roughness and crystallinity of the lm were signi cantly improved by using anisole as the antisolvent, the PCE of perovskite solar cells prepared with anisole as antisolvent reached 19.42%. [21,22] Based on the above research, our group proposed to apply environmentally friendly solvent to regulate the process of perovskite crystallization in order to achieve high-quality perovskite lms. We studied the effects of green organic solvents (such as propyl acetate and propylene glycol methyl ether) on the morphology and performance of perovskite solar cells. The results show that propyl acetate and propylene glycol methyl ether can promote the nucleation of PbI 2 particles and provide the heterogeneous nucleation site of perovskite crystal, the lms treated with propyl acetate and propylene glycol methyl ether have larger grains and lower RMS than those treated with chlorobenzene. The dense and full coverage perovskite lms with uniform grain size which is closer to the exciton diffusion length of perovskite can be obtained.
Compared with the CH 3 NH 3 PbI 3 perovskite soalr cells treated with chlorobenzene, the best e ciency of the devices treated with green antisolvent propylene glycol methyl ether is 21.60%, which has certain reference value and guiding signi cance for researchers to obtain environment-friendly high-quality perovskite solar cells in the future.

Results And Discussion
It is well known that chlorobenzene is widely used in the antisolvent method, and the perovskite lms are of good quality. Therefore, we choose the perovskite lm prepared by chlorobenzene as the control group.
First, the precursor solution of perovskite was dropped on the uoro doped tin oxide (FTO) glass substrate coated with a dense TiO 2 layer. The solution is then uniformly spread on the TiO 2 surface by rotating at a xed speed. In the process of spin coating, due to the slow volatilization of the solvent, the solute in the solvent will slowly separate out, and the color of the liquid lm will gradually change from bright yellow to dark yellow. At the moment of transformation, chlorobenzene is added to the crystal rapidly and uniformly. As mentioned before, the use of antisolvent can make the solution supersaturated and affect the nucleation of solute and the growth rate of crystal, and then affect the crystallization kinetics of perovskite materials and the nal morphology of the lm. When chlorobenzene drops on the wet lm, the lm immediately changes from transparent bright yellow to light brown. From the side, the lm is mirror like, after annealing, the lm is black. Then spin coating spiro ometad solution, then static annealing in the glove box for 24 h, and nally evaporate the Ag electrode. After the device is fabricated, its photovoltaic performance is tested. The process and sample of preparing perovskite lm by antisolvent are shown in Figure 1.
We systematically studied the effects of green organic solvents such as such as propyl acetate and propylene glycol methyl ether on the morphology control and performance of perovskite soalr cells. The SEM of the perovskite lms under different antisolvent conditions and the corresponding device parameters are measured in the experiment, as shown in Fig. 2 and table 1. It can be seen from Fig. 2 that when the perovskite precursor solution is spin coated and concentrated, anti solvent is dripped, which has different effects on the perovskite crystallization. The results show that the addition of antisolvent can promote the rapid crystallization of perovskite in different degree, so as to produce micron scale perovskite polycrystalline lms. Fig.2 (a) data shows that the grain size of the lm without any treatment is about 100nm~120nm, the uniformity is relatively poor and the surface roughness reaches 30nm. The series resistance of the corresponding device reaches 1000 Ω· cm 2 , and the parallel resistance is only 300 Ω· cm 2 , which is extremely unfavorable for the internal performance of the devices. Fig.2 (b) shows the lm treated with chlorobenzene. The particle size is about 120nm~150nm, the uniformity is improved,and the surface roughness is 25nm. This shows that the addition of antisolvent can effectively improve the crystallization speed and quality of the lm grain. Fig.2 (c) and (d) show that the high boiling point DMF and DMSO in the lm can be extracted quickly after dropping ether and 2-methyl-1-propanol can complete the rapid crystallization of perovskite at room temperature, and the grain size has been greatly improved, which can reach 200nm~250nm, which will be conducive to the separation and transmission of electrons and holes, so as to reduce the probability of the combination of the two and improve the short-circuit current density; however, due to the poor uniformity of the lm grain, this will objectively lead to the series resistance of the device or The larger one is more than 150 Ω· cm 2 . Fig. 2 (e) and (f) are perovskite lms treated by propyl acetate and isooctane, respectively. It is found that the grain size of the lm is about 330nm ~400nm, and the surface roughness is reduced to 12nm, showing a better the uniformity and roughness. The grain size and orientation growth of perovskite crystal are obviously improved. In particular, tetragonal phase is seen in Fig. 2 (e) and (f). We think that propyl acetate and isooctane have good solubility for methylamine salt in precursor, but poor solubility for PbI 2 . On the one hand, the antisolvent is conducive to the solubilization and transfer of methylamine salt particles, and promotes the mixing reaction between precursors. On the other hand, it can promote the nucleation of PbI 2 particles and provide the heterogeneous nucleation site of perovskite crystal, so as to promote the rapid growth of perovskite crystal. Therefore, the lms treated with propyl acetate and isooctane have larger grains, and lower RMS roughness than those treated with chlorobenzene. Larger grain size means less grain boundary and carrier recombination sites. In addition, lower roughness is conducive to the formation of better contact between perovskite lm and electron transport layer and the formation of larger short-circuit current density. Both propyl acetate and isooctane have obtained a denser calcium titanium ore bed with full coverage. This will inevitably lead to less grain boundaries between the crystals, thus reducing the charge recombination, greatly reducing the series resistance of the device (only 50 Ω· cm 2 ), while the parallel resistance reaches an amazing 3000 Ω· cm 2 , which means that a good ohmic contact and interface connection are formed inside the device. The absorption spectrum of CH 3 NH 3 PbI 3 lm when dropping different solutions of chlorobenzene and isooctane in the process of spin coating can be obtained, as shown in Figure 3(a). It can be seen that the absorption range of CH 3 NH 3 PbI 3 lm without any treatment is 500nm-800nm, the absorption strength is not large, and the grain size is basically below 130nm. We prepared CH 3 NH 3 PbI 3 lm with chlorobenzene and found that absorptive strength has been improved in the range from 500nm to 800nm. When isooctane was used as extractant, the absorption strength of the lm was increased by more than 55%. This is mainly because isooctane can regulate the grain size and orientation growth of CH 3 NH 3 PbI 3 .The grain size reaches 320nm and the grain boundary is signi cantly passivated. Therefore, the intensity of absorption can be improved. In addition, the anti-solvent can penetrate into MA + , Pb + and I -1 ion positions between perovskite molecules, regulate the narrowing of the band gap, and promote the red-shift of the absorption spectrum. We analyzed the XRD patterns of perovskite lms treaeded with chlorobenzene and isooctane, as shown in Figure 3(b). There are two small peaks at 11.8 ° and 12.9 ° for CH 3 NH 3 PbI 3 , which are the inactive δ phase of CH 3 NH 3 PbI 3 and the peak of the cubic PbI 2 phase. When we prepare CH 3 NH 3 PbI 3 lm with isooctane, the peaks of yellow phase and PbI 2 completely disappear.
The peaks of (100), (200), (210) and (220) will gradually increase, which fully shows that isooctane can promote the orientation growth of the lm, inhibit the phase formation of δ phase and cubic crystal PbI 2 , and obtain more stable perovskite crystal. Furthemore, we extrained the grain size from XRD data by the Scherer equation.We can get more accurate grain size for perovskite lms treaded with different antisovlents.The result shows that the grain size of the lm without any treatment is 105nm, and The particle size of the lm treated with chlorobenzene increases to 128nm. Dropping ether and 2-methyl-1propanol can complete the rapid crystallization of perovskite at room temperature, and the grain size has been greatly improved, which can reach 203nm and 236nm, respectively. It's even more obvious that this value has grown to 337nm and 389nm respectively after introducing propyl acetate and isooctane optimization. This fully shows that green antisolvent can effectively regulate the crystallization process of perovskite crystal and improve lm quality.
In order to further clarify the in uence of antisolvent on the photovoltaic performance of perovskite solar cells, we prepared perovskite solar cells with FTO/TiO2/perovskite / spiro-oMeTAD / Ag. High quality and high e ciency devices were obtained by adding different antisolvent in the preparation process. The measured data are shown in Fig.4(a) and table 2 respectively. It is found that PCE of perovskite solar cells without any green antisolvent treatment is only 14.50%, in which the open circuit voltage V OC is 1.10V, the short circuit current density J SC is 18.85mA/cm 2 and the lling factor FF is 69%. After chlorobenzene treatment, the device e ciency has been improved to 15.60%, the main reason is that the micro morphology of perovskite lm has been optimized, the grain size is larger, which makes the composite probability of hole and free electron decrease, resulting in the increase of short-circuit current density, reaching 20.26 mA/cm 2 . Further analysis shows that when we add different green antisolvent in turn, the effect on device e ciency is more obvious. For example, when we use ether or 2-methyl-1-propanol as antisolvent, the crystalline particle size of perovskite lm becomes large, reaching 350nm, which is more conducive to the dissociation and diffusion of holes and free electrons, so that the combination is reduced. The surface roughness is reduced to about 20 nm, The short-circuit current density J SC is increased to 25.10 mA/ cm 2 , the lling factor FF is increased to 72%, and the device e ciency after 2methyl-1-propanol treatment is up to 20.05%. Compared with the untreated perovskite solar cells, the energy conversion e ciency is improved by nearly 38%. Moreover, when propyl acetate and isooctane are added to the crystallization process of perovskite lm, the effect is more obvious. The antisolvent control of isooctane has achieved a dense and fully covered calcium titanium ore layer with uniform grain size, which is closer to the exciton diffusion length of perovskite, basically maintained at about 300nm, and the srface of the lm is smoother, only 12nm. These parameters show that there are less grain boundaries between the crystals, and the probability of charge recombination is greatly reduced. The series resistance of the devices treated with isooctane is only 50 Ω·cm 2 , while the parallel resistance is as high as 3000 Ω·cm 2 , which means that a good ohmic contact and interface connection are formed inside the devices, which makes the short-circuit current density J SC increase to 25.80 mA/cm 2 , the lling factor FF increased to 75%, and the optimal device e ciency increased to 21.60%. Furthermore, we found that the J-V curve obtained were basically consistent when using forward and reverse scanning to test the device e ciency, this may be due to the use of antisolvent green leads to the large perovskite grains and high quality perovskite thin lm, so as to effectively eliminate the ion migration at the interface and the non-radiative recombination, cause the hysteresis phenomenon is not obvious, as shown in Fig4 (b).

Conclusion
The results show that propyl acetate and isooctane can promote the nucleation of PbI 2 particles and provide the heterogeneous nucleation site of perovskite crystal, so as to promote the rapid growth of perovskite crystal. Therefore, the lms treated with propyl acetate and isooctane have larger grains and lower RMS roughness than those treated with chlorobenzene. Compared with the CH 3 NH 3 PbI 3 perovskite soalr cells treated with chlorobenzene (power conversion e ciency is 17.86%), the best e ciency of the devices treated with green antisolvent isooctane is 21.60%, which is improved by 21%. is prepared by adding liquid TiCl 4 to the frozen deionized water drop by drop. After that, the FTO immersed in TiCl 4 solution and the culture dish were sealed with fresh-keeping lm, then covered and placed in an oven at 70℃ for 1h. After the reaction, a large amount of deionization is used to wash until the solution is clear, and then the FTO substrate with TiO 2 is blown dry with a nitrogen gas gun and is annealed on a 150 ℃ hot bench for 15 min. Therefore, equal mass ratio of MAI and PbI 2 was dissolved in dimethyl formamide (DMF) and dimethylsulfoxide (DMSO) mixed solvent (volume ratio of 4:6), and the mass concentration of MAPbI 3 was 1.2 M/mL. Droping 100 μLof perovskite precursor solution on the surface of titanium oxide, spin coating at 1000 rpm for 10 s, then spin coating at 4000 rpm for 40 s, and droping 150μL chlorobenzene, propyl acetate and propylene glycol methyl ether , respectively, and transfered to 110 ℃ hot stage for annealing for 15 min. Next, dissolving 90 mg Spiro-OMeTAD in 1 mL of chlorobenzene, add 45μL of Li TFSI acetonitrile solution with a concentration of 170mg/mL, 75μLof kf209 acetonitrile solution with a concentration of 300 mg/mL and 10L μLof TBP, respectively. Spiro-OMeTAD layer was obtained by spin coating on the surface of perovskite lm at 4000 rpm for 20 s. At last, we vacuumize to 5x10 -4 Pa and control the evaporation speed within 1Å/s, and prepare the 80nm Ag electrode by vacuum evaporation.

Characterization of perovskite solar cells
The morphology of perovskite lms was observed by scanning electron microscopy (SEM, Nova Nano SEM 450, FEI, USA). The UV visible spectrometer (uv-3600, Shimadzu, Japan) was used to test the absorption of perovskite lms. Electrochemical impedance spectroscopy (EIS) measurements were conducted using Zahner Zennium electro chemical workstation in a frequency range from 0.01Hz to 1.5MHz. X-ray diffraction (XRD) patterns of perovskite lms were recorded by Rigaku D/MAX-2400 diffractometer. The J-V curves measurements of the devices were conducted by a Keithley 2611 sourcemeasure unit and a PVIV-201 VI-V Station (Newport Oriel). Its illumination source was calibrated by using a Si reference cell system (91150 V, Newport). The average power of a solar simulator was 100mW/cm 2 . All the J-V curves were obtained at a scanning speed of 200mV/s from 1.2V to-0.1V and from -0.1V to1.2V, respectively.
Declarations Figure 1 The processing of CH3NH3PbI3 perovskite lm by antisolvent  Photovoltaic performance (a) of CH3NH3PbI3 perovskite solar cells prepared by adding different antisolvent, and hysteresis characteristic (b) based on perovskite solar cell treaed with isooctane.