Preparation of methylammonium lead iodide (CH3NH3PbI3) thin film perovskite solar cells by chemical vapor deposition using methylamine gas (CH3NH2) and hydrogen iodide gas

An upscalable chemical vapor deposition setup has been built‐up and employed in producing methylammonium lead iodide (MAPI) thin film perovskite solar cells, leading to a maximum efficiency of 12.9%. The method makes use of methylamine gas and hydrogen iodide gas to transform a predeposited layer of lead(II)iodide (PbI2) into MAPI. Although the reaction mechanism includes the intermediate phases lead oxide (PbO) and lead hydroxide (Pb(OH)2), indicated at least on the surface of the samples by XPS, neither species could be observed in XRD measurements of the stepwise reaction, which show a mixture of highly oriented cubic and tetragonal MAPI perovskite lattice systems.


| INTRODUCTION
Since their emergence in 2012, 1,2 thin film organometal halide perovskite solar cells have experienced a uniquely fast improvement in certified efficiency of up to 25.2% in 2020. 3 Complementing silicon (Si) solar cells in tandem device structures, a certified efficiency of 29.1% has been achieved to date, 4 paving the way to reaching the current optimized efficiency of the best double-junction tandem solar cells of 32.9%, 5 that already overcomes the Shockley-Queisser limit of single-junction solar cells. 6 Challenges like reducing the environmental impact, increasing the stability, and reducing the cost of production even further remain. Popular routes to forming the photo-active perovskite methylammonium lead iodide (CH 3 NH 3 PbI 3 , MAPI) is either from solution, by spin coating, or from the gas phase, by evaporating the two precursor salts lead(II)iodide (PbI 2 ) and methylammonium iodide, (CH 3 NH 3 I, MAI). Since CH 3 NH 3 I can decompose to either ammonia (NH 3 ) and methyl iodide (CH 3 I), 7 or to methylamine (CH 3 NH 2 , MA) and hydrogen iodide (HI), 8 a chemical vapor deposition (CVD) route to forming MAPI perovskite would seem possible, by mixing either combinations of the decomposition precursors with lead(II)iodide. The standard route to synthesize the MAI salt is by mixing an ethanol, or methanol MA solution with hydroiodic acid, HI (aq.). [9][10][11] For further purification and recrystallization, a considerable amount of solvents like diethyl ether or isopropanol are used. Synthesizing MAPI from PbI 2 by directly using the two gases MA and HI with the respective boiling temperatures of −6°C and −35°C would thus be an alternative route. An industrially established process that could achieve this would thus be within reach, through the development of a highly upscalable CVD process.
Zhou et al 12 was one of the first to report the use of MA for healing MAPI perovskite films, describing the formation of a CH 3 NH 3 PbI 3 ·xCH 3 NH 2 liquid phase during the perovskite-gas interaction. Raga et al 13 reported shortly after a synthetic route to forming MAPI perovskite thin films that implies exposure of a PbI 2 film to an atmosphere of heated methylamine from ethanol on a hot plate, in combination with an atmosphere of heated hydroiodic acid (HI aq.). They achieved an efficiency of 12.7% with a sequential process and 15.3% with a simultaneous HI + MA process. The paper established a reaction mechanism between the MA, HI, and PbI 2 , as shown by the Equations 1-5. Water from the atmosphere is needed for the reactions in Equations 1 and 2 to take place. In the MA step, Equations 2 and 3, the formation of lead oxide (PbO) and lead hydroxide Pb(OH) 2 can be observed. These species would be regenerated to lead iodide (PbI 2 ) in the HI step, according to Equations 4 and 5.

| Glass coated with fluorine-doped tin oxide (FTO)
Pilkington NSG TEC15 FTO glass substrates have been used, cut as 2 cm × 2 cm squares, with a sheet resistance of 12-14 Ω/sq. and a thickness of 2.2 mm.

| c-TiO 2 layer
The compact TiO 2 layer is produced by spray pyrolysis on the glass/FTO substrates. Five hundred microlitre of titanium diisopropoxide bis(acetylacetonate), 75 wt. % in isopropanol (TIAA) from Merck, is mixed with 18 mL reaction grade Ethanol. This solution is sprayed using oxygen carrier gas onto the glass/FTO substrates that have been treated for 5 minutes in an oxygen plasma oven and heated to 450°C for 25 minutes prior to deposition. After the spraying process, the substrates are sintered for 30 minutes at 450°C in atmosphere.

| m-TiO 2 layer
The mesoporous TiO 2 layer is deposited by spin coating 100 µL of a 1:7 weight ratio solution of 18NR-T Titania (TiO 2 ) paste from Greatcell Solar and reaction grade ethanol onto each glass/FTO/c-TiO 2 substrates in atmospheric conditions. The solution is dropped on a substrate, before spinning at 83 rps (revolutions per second) for 45 seconds. After drying for 10 minutes at 70°C, an additional sintering step takes place for 45 minutes at 450°C on a hot plate, in atmospheric conditions.

| PbI 2 layer
Prior to the deposition of the lead(II)iodide layer, the glass/ FTO/c-TiO 2 /m-TiO 2 substrates are treated in a UV/ozone oven for 15 minutes. The deposition takes place in a nitrogen (N 2 )-filled glovebox. After each substrate has been heated for 2 minutes at 80°C on a hot plate, 100 µL of a 555 mg PbI 2 (Alfa Aesar 99.9985%, metal base) in 1 mL DMF (N,N-Dimethylformamide, Merck, 99.8%, anhydrous) solution that has been stirred for at least half an hour at 80°C is dropped onto the hot substrates, then spun at 108 rps for 90 seconds. Each substrate is dried for 10 minutes at 80°C.

| Gold layer
The front contact of the substrate is cleaned of the excess material through rubbing off with a cotton swab, dipped in isopropanol until transparent. The gold (Au) layer is deposited by argon sputtering in a Quorum Technologies Q300TD machine with 30 mA current for 120 seconds, using a steel mask.

| XRD setup
X-ray diffraction (XRD) patterns were recorded on a Bruker D8 Advance in Bragg-Brentano geometry with Cu Kα radiation and a VANTEC detector. The samples were measured in a 2θ-range between 10° and 90° using a fixed divergence slit of 0.3° for 60 minutes All samples were measured at ambient atmosphere, at most within few hours after production.

| XRD refinements
Analysis of diffraction data was performed using the Rietveld method with the program TOPAS V. 5.0 using the whole 2θ-range (10°-90°). The instrumental intensity distribution of the instrument was determined empirically from a sort of fundamental parameters set 27, using a reference scan of LaB 6 (NIST 660a) and silicon, respectively. The texture parameters required to fit the perovskite phases were calculated by application of the March model, 14 in relation to the (100) plane for the cubic phase, or the (100)(001) planes for the tetragonal model.

| XPS setup
We used a ThermoFisher Scientific ESCALAB 250X. Mg K α radiation with an excitation energy of 1254 eV and a take-off angle of 90°. The machine was calibrated using copper, gold, and silver standards.

| SEM setup
An FEI/Philips scanning electron microscope was used, of model FEX XL 30, that has a Schottky-type electron source and is equipped with an Everhart-Thornley secondary electron (SE) detector. The samples are tilted at 70°.

| PL setup
Varian Cary Eclipse, 350 nm excitation with Schott WG 11, 3 mm filter and Schott GG 450 nm as an emission filter.

| Solar simulator setup
The solar simulator is built by LOT Quantum Design, LSH102, using a USHIO UXL-150SO-Xenon Lamp and an AM 1.5G (air mass, global) filter. The samples were measured unmasked. The area of the cells was defined by the gold back contact.

DEPOSITION SETUP
The CVD setup used in this work allows the transformation of a lead iodide (PbI 2 ) thin film to MAPI (methylammonium lead iodide, CH 3 NH 3 PbI 3 ), in a multi-step cycled process, that involves the precursor gases methylamine (CH 3 NH 2 , short MA) and hydrogen iodide (HI). A photograph of the setup, built-up in a laboratory hood, can be seen in Figure 1. Figure 2 features a schematic drawing of its components. The main body consists of a three-zone tube furnace equipped with infrared heaters and a reaction glass tube with KF (klein flange) vacuum flanges. The temperature in each zone can be controlled by feedback loops of K-type thermocouples with feedthroughs inside the glass tube. Metal holders as seen in Figure 3 can hold up to three substrates, stacked vertically. In this setup, different types of gases can be inserted into the reaction tube, the pressure of which can be monitored using a Bourdon pressure gauge. Surrounding air or nitrogen is filled in by a manually controlled ball valve. Methylamine gas can be dosed using a standard steel lecture bottle with a manual pressure regulator. The HI gas is self-made using a glass apparatus connected to the reaction tube, by adding hydroiodic acid, mixed with a surplus of iodine, from a dropping funnel to red phosphorous. 15 The evolving HI gas is optionally passed through the drying agent phosphorus pentoxide (P 4 O 10 ) before entering the reaction tube, removing further water vapor from the formed gas, to promote the formation of the right side of Equations 4 and 5, according to Le Chatelier's equilibrium law. The complete setup can be evacuated to ca. 1 mbar with the help of a rotary vane pump. A manual membrane valve is used to disconnect the vacuum line from the reaction tube. To hinder condensation of the reactive gases inside the rotary vane pump, a cold trap with liquid nitrogen (LN 2 ) is connected between the reaction tube and the pump.

| CVD PROCEDURE
For transforming PbI 2 to MAPI, MA and HI are introduced in the reaction tube sequentially and in multiple cycles. A half cycle (0.5, 1.5, and 2.5) refers to a methylamine step, and a whole cycle (1.0 and 2.0) refers to an HI step. Prior to loading the spin-coated PbI 2 samples, the reaction tube is heated to 100°C for at least an hour under vacuum to remove any residuals that may adhere to the walls. All further steps take place at room temperature. For the methylamine step, 400 mbar of surrounding (humid) air is introduced into the reaction tube. After inserting an additional 200 mbar of methylamine, the yellow PbI 2 -coated substrates turn transparent, as seen in Figure 3.
Very small droplets can be observed on the surface that indicate a formed liquid phase of the methylamine and PbI 2 complex. [16][17][18] This intermediate phase is one of the advantages of the methylamine process, since the resulting film becomes very smooth, independent of the roughness of the initial precursor layer of PbI 2 , as shown by the SEM images in Figure S2. After 15 seconds, a 600 mbar air flow for 1 minutes removes the methylamine gas from the tube. In the first seconds of removing the methylamine gas, the substrates turn light brown, with gradual darkening. To remove all excess methylamine and air from the reaction tube, it is purged three times with nitrogen (1 bar to 1 mbar). In the next step, 200-300 mbar of HI gas is added to the evacuated tube. To produce the HI gas, 57 wt. % concentrated hydroiodic acid is mixed with solid iodine in a 1:2 weight ratio. The solution is dropped onto red phosphorous 15 and the resulting HI gas is optionally passed through the drying agent phosphorous pentoxide (P 4 O 10 ), to further remove excess water vapor. After 10 minutes reaction time, the HI gas is removed by purging three times with nitrogen. After evacuating, the methylamine and HI steps are repeated for 2.5 cycles.

| RESULTS AND DISCUSSION
The substrates of the intermediate steps of the CVD process, seen in Figure 4, have each been characterized by XRD- Figure 5 and UV/Vis- Figure 7. These samples were produced without the drying agent P 4 O 10 in the HI production step. Figure 5 shows that in all steps of the CVD process a high amount of oriented MAPI forms, having a (100) reflex at 14.2°, a (200) reflex at 28.5°, and a (300) reflex at 43.4°, which indicates strong texturing of the samples. These values are in accord with literature. 19,20 Although from Equations 2 and 3 the formation of PbO and Pb(OH) 2 is expected, neither species can be found in the XRD patterns of the samples; however, XPS measurements ( Figure S1) do indicate the formation of PbO and Pb(OH) 2 . These species merely form on the surface of the samples, only being detected by the surface-sensitive XPS technique, or they could exist as amorphous phases, or having very small crystallites spread out throughout the material, that could make them undetectable by the bulk measuring method XRD.
Rietveld refinements of the XRD measurements displayed in Figure 6 (2.5 MA sample) indicate the presence of a mixture of a cubic and a tetragonal MAPI perovskite phase. Rietveld refinements of the samples 1.0. HI, 1.5 MA, and 2.0. HI ( Figure S5-S7) also feature a mixture of the two phases. In contrast, the 0.5 MA sample from Figure S4 could be fitted with a cubic phase only. Results of the lattice parameters, phase fraction distributions, texture parameters, and volumes per ABX 3 unit are summarized in Table 1. The low values of the refined March-Dollase parameters, ranging from 0.16 to 0.21, indicate highly textured film growth with the (001) plane in parallel to the film surface. Two perovskite phases can be found within most of the films, indicated by the (400) reflection around 60° 2theta. These are high-volume phases, indicated to be predominantly tetragonal when found in high quantities, as well as a low-volume phase, predominantly cubic. 21 Their quantities show a strong dependence on the treatment step, where the low-volume phase becomes predominant after MA treatment.
The different phases with different volumes most likely indicate that the material can exist in different defect states with slightly different compositions. Figure 7 shows UV/Vis spectra of samples of each step of the CVD process. The optical band gap E g of the materials formed is red-shifted through the progressing cycles, from 1.61 eV (768 nm) for the PbI 2 layer that was treated only once with methylamine (0.5 MA), to 1.60 eV (773-774 nm) for the intermediate steps and finally 1.59 eV (778 nm) for the last step of the process (2.5 MA), indicating an increased amount of MAPI crystallinity. The band gaps are in accordance with literature values for MAPI. 22,23 The PL band of MAPI originates from a near-band-edge transition 24,25 and is with 1.63 eV (762 nm), for a 2.5 MA sample ( Figure S3), higher than the measured optical band gap of 1.59 eV (778 nm).  Solar cells of the CVD MAPI films use the stack sequence Glass/FTO/c-TiO 2 /m-TiO 2 /MAPI/spiro-MeOTAD/Gold, as illustrated in Figure 8, left. The substrates are 4 cm 2 in size, each holding four 32.5 mm 2 large cells. The cross-sectional SEM image of a completed cell, Figure 8, right, reveals a MAPI perovskite thick film of 400 nm, while the initial PbI 2 layer is 250 nm thick ( Figure S2). Figure 9 shows current-voltage curves of the champion cells under an AM1.5G illumination, measured with a scanning speed of 0.1 V/s, which were produced without and with the drying agent P 4 O 10 in the HI step, respectively. The best solar cell produced with the drying agent reached a 12.9% efficiency under MPPT (maximum power point tracking), while the one without the drying agent reached an efficiency of 11.7%, each for a 32.5 mm 2 cell on a 4 cm 2 substrate.
The spread of the efficiencies of the solar cells made with the drying agent (HI production step) is 1.9% with an average of 11.1%, and the spread for the cells made without the drying agent is 1.4% with an average of 9.4%. The individual values for the efficiencies, the open circuit voltages (V OC ), the short circuit currents (J SC ), and the fill factors for representative solar cells are given in Table 2. The low fill factors with an average value of around 60% may be related to PbO and Pb(OH) 2 species that have been found on the surface of the films by the XPS measurements presented in Figure S1.

| CONCLUSION
This work shows that an upscalable CVD process can be successfully employed to produce well-working thin film perovskite solar cells, with an efficiency of up to 12.9% for 2 × 2 cm 2 substrates and 32.5 mm 2 large cells, using a predeposited lead(II)iodide layer (PbI 2 ), methylamine gas (MA), and hydrogen iodide gas (HI). The method produces a mixture of highly oriented cubic and tetragonal perovskite material. One given limitation of the process is that the MAPI perovskite phase is nonpure, due to the nature of the mechanism, that involves water for the reaction and the byproducts lead oxide (PbO) and lead hydroxide (Pb(OH) 2 ). On the other hand, the process can be used in atmospheric conditions, using ambient, humid air. In addition, the process may be adapted to perovskite materials with more complex compositions.