Rapid Perovskite Formation by CH3NH2 Gas-Induced Intercalation and Reaction of PbI2

Solution processable perovskite solar cells traditionally use CH3NH3I solid powder as one of the two precursors that requires solvation into a solution and a spin-coating step; resulted films need post-annealing (~1h) for complete conversion to CH3NH3PbI3. Here we describe a method to form perovskite in ambient air by exposing PbI2 films to the simpler CH3NH2 gas precursor (as opposed to CH3NH3I solid powders) . The reaction completes within a few seconds forming complete-coverage perovskite films with a roughness of 2 nm. The non-stoichiometric reaction produces Pb oxides as by-product, which are reconverted by further HI gas exposures. With combined measurements of thin film crystal structure, chemical state, and absorption properties, we elucidate the chemical reaction mechanisms underlying these gas-induced processes. Fabricated solar cell devices show efficiency of 15.3%, which stays almost the same after 133 days. Such a gas-induced reaction also enabled the successful preparation of high quality perovskite films with a size of 100 cm. We also demonstrate that this method is versatile and enables the formation of a wide range of perovskite materials, e.g. CH3NH3PbCl3, CH3NH3PbBr3, CH3NH3PbI3, CH3NH3SnBr3 CH3NH3PbBrxI3-x, CH3NH3PbBrxCl3-x. Sonia R. Raga, Luis K. Ono, and Yabing Qi*, J. Mater. Chem. A, in press (2015) (DOI: 10.1039/C5TA10055K).


Introduction
2][3][4] With only a few years of research, perovskite solar cell efficiencies have rapidly increased to a level that is on par with even the best crystalline silicone solar cells. 5,6At present, solution processing is the most widely employed technique for the fabrication of perovskite layers. 7,8In solution processing, two solid powder precursors, methylammonium halides (CH 3 NH 3 X, X ¼ I and Br) and lead halides (PbX 2 , X ¼ I, Br, and Cl), are dissolved into solutions and spin-coated on substrates.A subsequent post-annealing step (typically $1 h) is required for the complete conversion to CH 3 NH 3 PbX 3 .Solution processing techniques, especially the related spin-coating technique, are difficult to scale up, which makes them unsuitable for massproduction of large-area solar cells.The industrial-scale manufacturing of perovskite solar cells urgently calls for a new method that is suitable to coat high-quality perovskite lms over a large area (e.g. 1 cm 2 or larger). 4n this work, we report a novel method that is capable of forming uniform perovskite lms in ambient air over a large area.It is achieved by exposing PbI 2 lms to a CH 3 NH 2 gas precursor (referred to as the MA step hereaer).Such a gasinduced reaction exhibits multiple advantages over the solution processing using the conventional CH 3 NH 3 I solid powder precursor.0][11][12][13] In general, 2D materials possess the characteristics of strong in-plane chemical bonds and weak van der Waals inter-plane interactions.The observation that CH 3 NH 3 PbI 3 perovskites are formed within merely a few seconds is a direct manifestation of the benet of the gas intercalation process owing to the high permeability of CH 3 NH 2 gas and the 2D layered structure of PbI 2 lms that enables CH 3 NH 2 to quickly (1) diffuse through grain boundaries of polycrystalline PbI 2 crystals and (2) subsequently to enter inside every PbI 2 crystallite.Secondly, because of the gas-induced reaction nature (as opposed to CH 3 NH 3 I solid powder), it allowed the successful preparation of high-quality perovskite lms with a size of 100 cm 2 .Thirdly, using different starting lead/tin halide lms reacting with CH 3 NH 2 as well as additional exposure to HI or HBr, our method enabled the formation of a wide range of perovskite materials (CH 3 NH 3 PbCl 3 , CH 3 NH 3 -PbBr 3 , CH 3 NH 3 PbI 3 , CH 3 NH 3 SnBr x I 3Àx , CH 3 NH 3 PbBr x I 3Àx , and CH 3 NH 3 PbBr x Cl 3Àx ).5][16][17] Fourthly, in comparison with solution processing, our method requires neither the synthesis of CH 3 NH 3 I nor post-annealing treatment resulting in a much simpler and faster process for perovskite fabrication (within 15 minutes), which helps reduce cost.Last but not least, using X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS) and UV-visible spectroscopy (UV-Vis), we propose the chemical reaction mechanisms underlying these gas-induced reactions that are in good agreement with experimental observations.Experimental PbI 2 layer fabrication PbI 2 was dissolved in dimethylformamide (DMF, 460 mg mL À1 ) and kept stirring at 70 C until it completely dissolved.The PbI 2 solution was spin coated on the substrates (glass or FTO/ compact layer TiO 2 /mesoporous TiO 2 ), previously heated at 70 C, at 4000 rpm for 30 s and then dried on a hot plate at 70 C for 5 minutes in a N 2 glovebox.The substrates were taken out to air for perovskite formation by subsequent MA and HI steps as well as Processes P1 and P2 described below.

MA step
Methylamine gas exposure step.In order to grow perovskite from the PbI 2 substrates, a small Petri dish with an opening of 2 cm was lled with $200 mL of methylamine (MA) solution in ethanol (33% wt, Aldrich) and placed on a hot plate at 45 C. The PbI 2 substrate was placed at the bottom of an 8 cm tall beaker using double sided tape.The beaker was placed upside down covering the MA-containing Petri dish, with the PbI 2 substrate facing the vial opening.A schematic illustration of the setup is shown in the lower central region in Fig. 1.Aer 5 s, the PbI 2 lm starts changing to a brownish color, and aer 5 s additional MA exposure, the lm turns transparent (Fig. 1, the lm with excess CH 3 NH 2 ).At this point, once the beaker with the substrate is removed from the MA gas, the transparent substrate turns brown within 5 s (Fig. 1, intermediate stage).A video clip was provided as ESI.† All the samples passed through these entire color changes happening within one minute ensuring that the reaction reached the bulk of the lm (saturation conditions).This procedure was named the MA step in this work.The postannealing studies were performed aer PbI 2 lms were saturated with MA gas (i.e.transparent lm was observed), and the substrates were placed immediately on a second hot plate preheated at 45, 70 or 100 C, and le for two minutes.

HI step
Hydroiodic acid gas exposure step.The perovskite lms prepared by the MA step were subsequently exposed to HI to convert lead oxide to PbI 2 .The substrates were placed on a hot plate at 70 C and covered by a glass Petri dish of 5 cm in diameter.The HI exposure was performed by dropping some HI (57%, TCI co.) $50 mL in a small crucible inside the Petri dish and next to the samples.The HI gas lled the space inside the Petri dish, reacting with the lm.The exposure times varied from 2 to 10 min and were indicated in the main text.Optimized HI exposure time based on XRD, UV-Vis, and XPS was between 2 and 4 min.

Process P1
Sequential MA gas and HI gas exposures.The substrates were placed on the hot plate and covered by a glass Petri dish of 5 cm in diameter.A volume of $200 mL of MA solution or $50 mL of HI was placed in a crucible next to the substrates on the hot-plate at 70 C.For the best results, the treatment to PbI 2 substrates was initiated rst with the MA exposure step as described above.The optimized treatments were HI exposure for 2 min and MA exposure for 4 min, venting the remaining gas before changing the material.Contrary to the rst MA step, the lm did not turn transparent in this second or further MA steps aer 4 min.

Process P2
Simultaneous HI gas and MA gas exposures.For the simultaneous HI and MA exposures, the substrates were placed on the hot plate at 70 C and covered by a glass Petri dish of 5 cm in diameter, with a small gap of $2 mm between the hot plate surface and Petri dish to let the air ow inside the Petri dish.For the simultaneous HI and MA exposures, small crucibles lled with HI $ 100 mL and MA solution $ 400 mL were placed inside the Petri dish and next to the samples.In this reaction, a white-dense homogeneous smoke was formed induced by the air ow below the Petri dish.We identied the deposited powder as methylammonium iodide by XRD.The substrates were le for 10 min and the surface was blown with N 2 gas.

TiO 2 substrate fabrication
Glass coated with uorine-doped SnO 2 (FTO, 15 Ohm sq.À1 ) was rst etched using Zn and HCl and cleaned by sequential sonication in a soapy bath, Milli-Q water and isopropanol.The FTO was covered with a 70 nm compact layer of TiO 2 by a spray pyrolysis method using a precursor solution composed of acetylacetone, Ti(IV) isopropoxide and anhydrous ethanol (3 : 3 : 2 wt).A 150 nm-thick TiO 2 mesostructured (mp-TiO 2 ) layer was deposited by spin-coating a diluted TiO 2 paste (Dyesol, 90-T) and then annealed at 450 C for 30 min.

Solar cell fabrication
The perovskite was deposited on the mp-TiO 2 following MA and HI steps and Processes P1 and P2.The hole transport layer was deposited by spin coating at 2000 rpm a chlorobenzene solution containing 59 mM 2,2 0 ,7,7 0 -tetrakis(N,N 0 -di-p-methoxyphenylamine)-9,9 0 -spirobiuorene (spiro-OMeTAD), 172 mM tertbutylpyridine and 32 mM of Li-bis(triuoromethanesulfonyl)imide salt pre-dissolved in acetonitrile.The top contact was fabricated by thermal evaporation of gold, with a nal thickness of 80 nm.The devices were not sealed.

Characterization
The lm crystallinity was analyzed by XRD (D8 Bruker) with a diffraction angle resolution of 0.01 .UV-Vis absorbance measurements were taken with a spectrophotometer Evolution 600, Thermo Scientic at intervals of 0.2 nm.The measured relative humidity and temperature of the laboratory were 50% and 25 C, respectively.The surface morphology was measured by using an atomic force microscope (AFM) from Asylum instruments using tapping mode cantilevers with a nominal spring constant of 48 N m À1 .For high resolution XPS (HR-XPS) measurements, we used a monochromated Al-Ka (hn ¼ 1486.6 eV) excitation source.The system comprises a delay-line detector with a 180 hemispherical, 165 mm mean radius analyzer, AXIS Ultra DLD (Kratos Surface Analysis).Binding energy scales for XPS were calibrated by measuring the Fermi edge (E F ¼ 0 eV) and Au-4f 7/2 (84.0 eV) on a clean Au surface.Energy resolution was less than 0.5 eV for HR-XPS.X-ray induced sample damage was monitored by taking ve consecutive spectra and comparing them.Acquisition time for each scan varied from 10 to 20 s, depending on the core level regions.The ve scans were averaged to a single spectrum if no changes were observed among them.Analysis of XPS data was performed in CasaXPS 2.3.16soware.Shirley-type background was subtracted to account for inelastic scattering processes.The atomic concentrations of different chemical elements were extracted from XPS intensity ratios aer appropriate normalization using atomic sensitivity factors.
Current-voltage (j-V) characteristics of solar cells were measured under 1-sun illumination (AM 1.5G, 100 mW cm À2 ) in air using a xenon-lamp (USHIO, UXL-150SO) based solar simulator (Newport Oriel Sol 1A) and a Keithley 2420 source meter.The solar simulator illumination intensity was calibrated using a monocrystalline silicon cell (Oriel Instruments Model Number 90026564, 2 Â 2 cm 2 ) calibrated by NREL.Inuences of voltage scan conditions and pre-illumination were tested and described in Fig. S9.† Unless stated, the voltage scan was performed from 1.2 V to 0 V (named reverse scan) at 0.1 V s À1 , with a dwell time of 5 ms and pre-illumination of 5 s.All measurements were performed in air at room temperature and cells without encapsulation.The effective active areas of the devices ranged between 0.05 and 0.18 cm 2 , and all measurements were performed without a mask.The external quantum efficiency (EQE) was measured using an Oriel IQE-200 measurement system from 850 to 300 nm in a 10 nm step.

Results and discussion
Detailed procedures are provided in the Experimental section.Briey, a PbI 2 lm spin-coated on a FTO/compact layer TiO 2 / mesoporous TiO 2 substrate was exposed to CH 3 NH 2 gas.The CH 3 NH 2 gas-exposed PbI 2 lm is named Sample #1 and this process is named MA step (Fig. 1 and Table 1) hereaer.A simple setup, shown in Fig. 1, was used for conning the CH 3 NH 2 gas inside a beaker with the substrate affixed at the bottom inside a beaker and facing upside down.The CH 3 NH 2 gas was generated by heating the CH 3 NH 2 solution at 45 C using a hot plate.The MA step completes within a few seconds generating a dark-red/brown color with semi-transparent hue, similar to that of CH 3 NH 3 PbI 3 perovskite lms synthesized by spin coating using CH 3 NH 3 I and PbI 2 precursors. 18Using the MA step, we were able to form a high-quality mirror-like semitransparent perovskite lm on a large substrate with a size of 100 cm 2 (ESI Fig. S1 †) in a few seconds, clearly demonstrating the potential of this method for large-area solar cell mass-production.We propose that because of the small size and high vapor pressure (350 kPa, 25 C) of CH 3 NH 2 gas molecules, they easily permeate into the PbI 2 lm.
The resulting lms of Samples #1 were extremely uniform and smooth with a mirror-like appearance.The smoothness was further conrmed by atomic force microscopy (AFM) measurements (ESI Fig. S2a and b †).The morphology of the initial PbI 2 lms showed a root mean square (RMS) surface roughness of $21 nm, that became $10 times smoother aer the MA step.The RMS roughness of Samples #1 was determined to be $2 nm, which explains the mirror-like pinholefree appearance.An amorphous, liquid phase has been observed previously upon perovskite crystal exposure to MA gas, which improves the perovskite lm coverage and surface roughness. 19Despite the smoothness of the lms, the crystalline features (grain sizes) in Samples #1 were extremely small and in the range of 40-70 nm.Solar cell devices employing the lms synthesized by only the MA step (i.e.no further treatments) as absorbers with subsequent stacked layers of spiro-OMeTAD and Au as a hole transport layer (HTL) and electrode, respectively, generated power conversion efficiencies (PCEs) as high as 7.6% with a short-circuit density (j sc ) of 14.2 mA cm À2 , an open-circuit voltage (V oc ) of 1.02 V, and a ll factor (FF) of 0.52 (Fig. 2f and Table 1).As will be described in detail later, the major composition of the lms aer the MA step is indeed stoichiometric CH 3 NH 3 PbI 3 perovskite.However, additional lead oxide and Pb(OH) 2 phases were also observed to form and distributed in the lms as a result of the MA step.The lead oxide and Pb(OH) 2 phases were formed as a byproduct based on the proposed reaction (1a) and the alternative reaction pathways (1b and 1c), 20 which are in good agreement with our experimental results of the thin lm crystal structure, chemical state, and absorption properties.
Table 1 Sample names and corresponding treatments and solar cell parameters obtained from perovskites formed by the different processes.The average power conversion efficiency (PCE), short-circuit current (j sc ), open-circuit voltage (V oc ), and fill factor (FF) values were calculated based on statistical analysis from at least 6 different cells.Solar cell parameters corresponding to the cell with best performance are indicated in brackets.Sample #3 corresponds to the device after 3 cycles of Process P1   The fact that the MA step can proceed even at room temperature (ESI Fig. S3 †) clearly suggests its spontaneity, i.e. the formation of stoichiometric perovskite and PbO/Pb(OH) 2 is thermodynamically more favorable as opposed to both (1) the reactants PbI 2 , MA and H 2 O and (2) the formation of non-stoichiometric perovskite containing a large number of iodine vacancies (i.e.PbI 2 + MA / MAPbI 2 ).We further introduced additional post-treatments (Fig. 2a-d) of sequential HI gas exposure (the HI step generating Samples #2), sequential HI gas and MA gas exposure (Process P1 generating Samples #3), and simultaneous HI gas and MA gas exposure (Process P2 generating Samples #4), which have shown to be efficient methodologies to convert the additional lead oxide and Pb(OH) 2 phases to PbI 2 and perovskite (Fig. 2a-d).If the number of cycles is not specied, Process P1 refers to a single sequential cycle of HI gas and MA gas exposure.Solar cells prepared by the HI step showed the best efficiency of 8.9% with a V oc of 1.05 V, a j sc of 12.5 mA cm À2 , and a FF of 0.68 (Fig. 2f and Table 1).In particular, the large increase of the FF compared to Samples #1 conrms better charge transport properties of PbI 2 than lead oxide and Pb(OH) 2 . 21The slight decrease in j sc (14.4 mA cm À2 / 12.1 mA cm À2 ) may be correlated with some perovskites being reconverted to PbI 2 in the HI step lowering the photocurrent generation.In contrast to the slight efficiency improvement by the HI step, both Processes P1 and P2 led to substantial improvement in solar cell performance, with the best cell achieving an efficiency of 15.3% with a V oc of 1.05 V, a j sc of 20.6 mA cm À2 , and a FF of 0.71 (Fig. 2f and Table 1).The device retained an efficiency of 14.6% aer >3000 h stored in a glove box (ESI Fig. S8 †), with an integrated j sc of 17.0 mA cm À2 extracted from external quantum efficiency (EQE) and a steady current at a peak point of 16.5 mA cm À2 (ESI Fig. S4 †).These processes (Processes P1 and P2) are fast, simple (without the need to synthesize MAI), and low-cost and can be easily scaled up to mass-production of large-area cells.
The physico-chemical properties of the synthesized lms from the MA and HI steps as well as Processes P1 and P2 (corresponding to Samples #1-#4, respectively) were investigated in detail by XRD (Fig. 2e), UV-Vis (ESI Fig. S2e †), XPS (ESI Fig. S2f †), and AFM (ESI Fig. S2a-d †).XRD measurements (Fig. 2e) performed on the PbI 2 lm before and aer the MA step conrmed the conversion to the tetragonal perovskite crystal with no residual PbI 2 . 22,23The XRD peaks at 14.1 and 28.5 peaks matched well with the reported (110) and (220) crystal peaks, respectively, of stoichiometric CH 3 NH 3 PbI 3 perovskite lms prepared by the standard method, i.e. the reaction between PbI 2 and CH 3 NH 3 I. 23 The UV-Vis absorption data (ESI Fig. S2e †) showed a sharp absorption edge at 775 nm corresponding to a bandgap of 1.6 eV in good agreement with the previous reports on CH 3 NH 3 PbI 3 perovskite. 24We also studied the effect of post-annealing on the crystallinity of Samples #1 formed at 24 C, the results of which can be found in the ESI Fig. S3 †.Our analysis indicates that the perovskite crystals formed in Samples #1 are stoichiometric CH 3 NH 3 PbI 3 and did not have an appreciable number of iodine vacancies.This result seems counter-intuitive that the MA step involving only PbI 2 and CH 3 NH 2 can form stoichiometric CH 3 NH 3 PbI 3 perovskite.On the other hand, the reaction underlying the MA step can be hinted at by our observation that perovskite can only be formed by performing the MA step in ambient air (i.e. in the presence of H 2 O) as described in reaction (1a).To highlight the importance of H 2 O in the reaction, a control experiment of the MA step was performed in a dry environment such as in a N 2 glovebox with H 2 O levels below 0.1 ppm (ESI Fig. S5 †).No perovskite formation was observed during this control experiment, which is in good agreement with the previously reported study performed with a vacuum system. 11PS analysis of the resulting lms from the MA step conrmed the formation of lead oxide and Pb(OH) 2 phases on Samples #1, which is consistent with the two proposed reaction pathways (1a) and (1b) + (1c) (ESI Fig. S6a and c † -O 1s corelevel regions).In fact, only Samples #1 showed the presence of two oxidation states in the XPS O 1s core-level region.6][27] More than twenty lead oxide compounds were reported to exist without a dened stoichiometry.This is consistent with the fact that the full width at half maximum (FWHM) value of peak O2 (1.9 eV) is signicantly larger than that of peak O1 (FWHM ¼ 0.6 eV).In those compounds, the crystallographic valence of the oxygen in the lattice adopts a wide range of values (O À1.89 to O À2.25 ). 27Due to such variations, it is challenging to determine the crystal and electronic structures of these lead oxide compounds.For the sake of simplicity, in the discussion hereaer, we will use PbO as an example of the product in (1a) to illustrate, even though other Pb x O y products are also very likely formed in Samples #1.XPS atomic ratios normalized with respect to Pb (ESI Fig. S2f †) were calculated based on the peak areas and with appropriate normalization taking into account atomic sensitivity factors.On the basis of the [O]/[Pb] and [I]/[Pb] values, the total amount of lead oxide and Pb(OH) 2 was calculated to be $24 mol% that coexists with perovskite in Samples #1.This is in good agreement with the proposed reactions (1a) and (1c), in which a nominal $33 mol% of lead oxide and Pb(OH) 2 would be expected.
The corresponding XRD results from Samples #2 (Fig. 2e) reveal the formation of PbI 2 crystals from the transformation of lead oxide and Pb(OH) 2 phases following chemical reactions (1d) and (1e) or by the iodination process due to the co-existence of I 2 gas in the HI fume. 28The peaks in the XPS O 1s core level region in Samples #1 became much smaller aer the HI step (2-10 min) (ESI Fig. S2f, S6b and S7 †), which corroborates the proposed reactions (1d) and (1e).Extra care was taken to minimize the exposure of freshly prepared samples to ambient air when transferring to the XPS measurement system.We excluded the possibility that oxygen signals in Samples #1 can come from contamination or water adsorption because they are present only in Samples #1, but not in Samples #2.The iodine incorporation can also be reected by an overall increase of the optical absorbance of the lms below $775 nm, which was observed to saturate aer performing the HI step for $4 min (ESI Fig. S2f and S7 †).Finally, much purer CH 3 NH 3 PbI 3 perovskite was formed aer the subsequent Processes P1 and P2, which is conrmed by XRD with no residual PbI 2 (Fig. 2e).XPS conrmed the perovskite lm formation.The saturated [I]/[Pb] ratio was close to 2.5 (ESI Fig. S2f †).The deviation from the expected [I]/[Pb] ratio of 3 could be attributed to several factors such as (i) XPS is surface sensitive ($5 nm) and the surface chemical composition can be different from the bulk chemical composition.Possibly for a similar reason, we observed the [N]/ [Pb] ratio also deviating from 1; (ii) the Pb 4f region shows convoluted Pb 2+ states from perovskite and lead oxide and Pb(OH) 2 .Thus, the [I]/[Pb total ] shows a smaller ratio.Indeed we noticed that some amount of lead oxide and Pb(OH) 2 was formed aer each cycle in Samples #3 (ESI Fig. S6c †).However, the amount is reduced rapidly aer each cycle of Process P1.On the basis of chemical reactions (1a) and (1c), aer each cycle of Process P1 the lead oxide phase is expected to reduce by $67%.For instance, 3 cycles of Process P1 are sufficient to bring the molar concentration of lead oxide phase down to 3.6 mol%.This trend of reducing PbO is well aligned with the increased PCE with the number of Process P1 cycles performed (ESI Fig. S8 and Table S1 †).In addition, we performed AFM measurements to study the crystal grain size and observed that the sequential and complete cycles of Process P1 and simultaneous MA and HI exposure of Process P2 in air produced a gradual increase of the CH 3 NH 3 PbI 3 crystal grains up to $200-400 nm (ESI Fig. S2c and  d †).
In addition, our method is found to be universal forming a wide variety of perovskite materials (CH 3 NH 3 PbCl 3 , CH 3 -NH 3 PbBr 3 , CH 3 NH 3 PbI 3 , CH 3 NH 3 SnBr 3 CH 3 NH 3 PbBr x I 3Àx , and CH 3 NH 3 PbBr x Cl 3Àx ).For example, we performed the MA step to convert the initial PbBr 2 and PbCl 2 lms into CH 3 NH 3 PbBr 3 and CH 3 NH 3 PbCl 3 perovskites, respectively (Fig. 3a).It was found that the time scale for CH 3 NH 2 intercalation into the Pb halide was much slower for the PbCl 2 ($180 s) than PbBr 2 ($60 s) and PbI 2 ($15 s) suggesting that the formation of iodine-based perovskite is more favorable than the Br and Cl counterparts.When the MA step-treated PbCl 2 (CH 3 NH 3 PbCl 3 ) lm was exposed to HI, only XRD peaks corresponding to CH 3 NH 3 PbI 3 could be observed. 18Furthermore, we were able to generate mixed-perovskite halides using our method by re-converting the $33% PbO and Pb(OH) 2 formed by the MA step by exposing to HI or HBr gas.For example, a PbBr 2 lm initially exposed to MA gas was subsequently exposed to HI gas in a stepwise manner inducing the gradual formation of CH 3 NH 3 PbBr 3Àx I x perovskite.Fig. 3b shows the gradual band-gap tuning of the CH 3 -NH 3 PbBr 3Àx I x perovskite lm.Long exposure times caused the optical absorption edge to saturate at 690 nm.We attributed this saturation to the formation of CH 3 NH 3 PbBr 2 I perovskite, where iodine substituted all of the 33% PbO and Pb(OH) 2 formed during the MA step.The XRD data (Fig. 3c) for the mixed Br : I perovskite showed a shi in the peak position, indicating that Br-and I-based perovskites are mixed rather than segregated.This suggests that HI gas permeates easily into the bulk of the lm generating uniformly mixed Br : I perovskite.In contrast, the CH 3 NH 3 PbCl 3 lm with PbO and Pb(OH) 2 exposed briey to HBr led to the appearance of two XRD peaks at 15 and 15.6 , which were assigned to the two segregated phases of CH 3 NH 3 PbBr 3 and CH 3 NH 3 PbCl 3 perovskites, respectively. 18In the case of SnBr 2 , the conversion was not possible in air because the formation of SnO and CH 3 NH 3 Br was more favorable.When we performed the MA step in a N 2 glovebox followed by HI gas exposure, the conversion to CH 3 NH 3 SnBr x I 3Àx perovskite was successful, but a certain amount of SnI 2 was also generated additionally.

Conclusions
We showed the role and interplay of three simple precursors, i.e.PbI 2 , methylamine (CH 3 NH 2 , MA) gas, and hydroiodic (HI) gas for the formation of high quality perovskite lms.MA gas (instead of the well-known precursor CH 3 NH 3 I), in ambient air, reacts directly with PbI 2 to form stoichiometric CH 3 NH 3 PbI 3 perovskite (the MA step).Because of the high permeability of the MA gas and the 2D layered structure of PbI 2 , the reaction completes within a few seconds.The MA step also leads to the formation of a relatively large amount of lead oxide (33 mol%) as a byproduct.Nevertheless, even with such a large amount of lead oxide, solar cells based on the MA step reached an efficiency of 7.6%.We further introduced post-treatments such as HI gas exposure, sequential and simultaneous HI and MA gas exposures.The post-treatments led to substantially improved performance with a negligible content of PbI 2 or Pb-oxides, with the best cell achieving efficiency of 15.3% and outstanding stability aer 133 days.

Fig. 1
Fig. 1 Methylamine gas induced perovskite formation.Illustration of the setup used for processing perovskite at room temperature from simple precursors of CH 3 NH 2 gas and the PbI 2 film is depicted in the lower middle part of the figure: a PbI 2 film pre-deposited on mp-TiO 2 /cl-TiO 2 /FTO is exposed to CH 3 NH 2 gas from a heated solution.The different stages of the reaction are schematized from left to right with an arrow corresponding to the time scale of the process.Upon CH 3 NH 2 gas exposure, the characteristic yellow colored PbI 2 film turns dark after 5 s.At this point, CH 3 NH 3 PbI 3 perovskite, PbO and Pb(OH) 2 are formed.After subsequent 10 s exposure, the film becomes transparent and associated with the incorporation of excess CH 3 NH 2 into the lattice.Once the sample is removed from the CH 3 NH 2 exposure, the film becomes dark red/brown in color within 5 s with a mirror-like surface.

Fig. 2
Fig. 2 Illustration of post-treatments of HI exposure and Processes P1 and P2 and corresponding solar cell performances.Overview of all the post-treatment processes is described in the text.(a) Exposure of the PbI 2 film to CH 3 NH 2 gas at room temperature (25 C) generates perovskite and PbO/Pb(OH) 2 with characteristic grain sizes in the range of 40 nm and a RMS surface roughness of $2 nm (Sample #1the MA step).(b) Sample #1 exposed to HI gas at 70 C re-converts the PbO/Pb(OH) 2 to PbI 2 (Sample #2the HI step).(c) Sample #1 exposed sequentially to MA and HI gradually converts the PbO/Pb(OH) 2 first to PbI 2 and eventually to perovskite crystals of $400 nm in size and $14 nm surface roughness (Sample #3 -Process P1).(d) Sample #1 exposed simultaneously to MA and HI at 70 C forms a smooth perovskite film with a grain size of $200 nm and roughness of $6 nm.(e) XRD and (f) j-V curves of solar cells fabricated based on films (a-d).The measurements were conducted under AM 1.5 (100 mW cm À2 ) illumination conditions (see the Experimental section for more details).

Fig. 3
Fig. 3 Perovskite conversion from PbX 2 (X ¼ I, Br, and Cl) films exposed to CH 3 NH 2 gas and to HI and HBr gases.(a) Scheme representing all the possibilities of lead halide exposures to CH 3 NH 2 and hydrogen halide gases.After the initial conversion of PbX 2 to CH 3 NH 3 PbX 3 + PbO/Pb(OH) 2 upon CH 3 NH 2 gas exposure, the perovskite can be finely tuned to mixed-halide perovskite, by converting PbO/Pb(OH) 2 to PbI 2 or PbBr 2 after HI or HBr gas exposures, respectively.The reactions happen at different time-scales depending on the halide: I (15 s) < Br (60 s) < Cl (180 s) with iodine-based perovskite conversion the fastest among all.Addition of chlorine is not favorable to form uniformly mixed-halide perovskite, but only phase-segregated perovskites.(b) Absorbance spectra and (c) XRD of selected films of mixed-halide perovskites.