A Generalized Crystallization Protocol for Scalable Deposition of High‐Quality Perovskite Thin Films for Photovoltaic Applications

Abstract Metal halide perovskite solar cells (PSCs) have raised considerable scientific interest due to their high cost‐efficiency potential for photovoltaic solar energy conversion. As PSCs already are meeting the efficiency requirements for renewable power generation, more attention is given to further technological barriers as environmental stability and reliability. However, the most major obstacle limiting commercialization of PSCs is the lack of a reliable and scalable process for thin film production. Here, a generic crystallization strategy that allows the controlled growth of highly qualitative perovskite films via a one‐step blade coating is reported. Through rational ink formulation in combination with a facile vacuum‐assisted precrystallization strategy, it is possible to produce dense and uniform perovskite films with high crystallinity on large areas. The universal application of the method is demonstrated at the hand of three typical perovskite compositions with different band gaps. P‐i‐n perovskite solar cells show fill factors up to 80%, underpinning the statement of the importance of controlling crystallization dynamics. The methodology provides important progress toward the realization of cost‐effective large‐area perovskite solar cells for practical applications.


Introduction
Deposition of photoactive absorbers by scalable printing methods is an attractive approach to realize the cost potential of emerging photovoltaic technologies toward industrial these printing methods, doctor blading is one of the most promising lab techniques because it is easy to handle with minimized material waste and, importantly, it can be readily transferred to slot-die coating for sheet-to-sheet or roll-to-roll volume manufacturing. Blade coating can be carried out at elevated substrate temperatures which has been maturely adopted to prepare organic optoelectronic devices, [18,19] but, is limited in controlling the ink temperature. Recently, several groups have reported progress in blade coating highly qualitative perovskite films at substrate temperatures over 100 °C. [9][10][11][12][20][21][22] However, it has been noticed that elevated processing temperatures close to or even higher than the temperature required for perovskite crystallization can lead to simultaneous solvent evaporation and crystal growth which, in turn, would constrain the processing window as well as the repeatability of device fabrication. In this context, development of a controlled crystallization protocol that is compatible with low-temperature deposition of precursor films is highly demanded for large-scale perovskite thin film manufacture.
Crystallization is a complicated process which is aimed at creating a crystalline material from liquid, gaseous, or amorphous solid systems. Nucleation and crystal growth are the two fundamental steps of vital importance which usually take place during the formation of a supersaturated state. Specific for polycrystalline perovskite thin-film growth, it generally involves three steps: i) deposition of a liquid precursor film onto a substrate, ii) drying of the wet film to reach a supersaturation followed by nucleation, and, iii) thermal annealing, where necessary, to facilitate perovskite crystal growth. The paramount challenge to realize high-quality perovskite films via printing methods is that the two critical steps, nucleation and crystal growth, usually take place in seconds and often simultaneously, making perovskite crystallization intractable. In the classical antisolvent crystallization strategy for spin-coated perovskite films, it is commonly recognized that a temporal intermediate phase is created upon applying an immiscible solvent at a critical time of high-speed substrate rotation. [6] This frozen intermediate stage alleviates the otherwise fast nucleation and crystal growth, leading to uniform and dense perovskite films upon a thermal annealing. Despite the ground success of spin-coating for processing perovskite films in the lab, the well-developed antisolvent strategy is extremely complex to transfer to large area printing lines due to the excess and spilling of washing solvents.
In this paper, we report a generic crystallization strategy that allows depositing high-quality perovskite thin films by blade coating at ambient temperature. We highlight that the key to the successful implementation of the technology relies on a combination of judiciously engineered precursor formulations and a deliberately designed intermediate crystallization step. Specifically, incorporation of a small amount of additive methylammonium chloride (MACl) into the perovskite precursor not only allows for fine tuning the morphology and crystallinity of the film but, more importantly, the presence of MACl can effectively suppress the formation of nonperovskite δ-phase in the formamidinium and cesium (FACs)-involved material system. Another central design of our protocol is the deployment of a vacuum-based precrystallization procedure, which is decisive to create intermediate phases by removing the excess solvent of the freshly printed films. The resulting intermediate stage guarantees that the precursor deposition is kept separated from the subsequent thermal-annealing based crystallization, leading to the growth of high-quality perovskite films in a controlled manner. To demonstrate the universal application of the protocol, we prepared three perovskite material systems with exemplary compositional cations and halides including: MAPbI 3 (1.58 eV), FA 0.95 Cs 0.05 PbI 3 (1.52 eV), and FA 0.6 MA 0.4 Pb(I 0.6 Br 0.4 ) 3 (1.76 eV). All three compositions are demonstrated with blade coating to reach a device performance on par with spin coating. In particular, a high fill factor (FF) up to 80% (obtained in 0.09 cm 2 device) and large-area module (4 × 4 cm 2 ) with negligible potential losses demonstrated the high quality of the perovskite films and the reliability of the crystallization protocol.

Results and Discussions
To underline the significance of this work, we first compare the "efficiency" versus "precursor deposition temperature" for the scalable fabrication of perovskite solar cells from previous reports. The collected data points are plotted in Figure 1 with the specific values and the corresponding references summarized in Table S1 (Supporting Information). From the plot, we can see that most of the perovskite solar cells were deposited at temperatures over 100 °C. As mentioned above, perovskite crystallization can take place concurrently with the solvent evaporation during precursor deposition at such high temperatures, which makes it challenging to control the crystal morphology. This deficiency will be addressed in this work by decoupling the precursor deposition with the subsequent thermal annealing via additive-and vacuum-assisted crystallization strategy. Our technology ultimately provides a simple yet reliable crystallization protocol for blade deposition of perovskite solar cells with a wide range of material composites.

Precursor Ink Deployment and Film Characterizations
We start with a rational design of perovskite precursor ink by means of additive engineering. Previous works have shown that chemical additives such as organic molecules, [23][24][25][26] inorganic salts, [27] and ionic liquids [28] etc. incorporated into perovskite precursors can have significant influence on the crystal quality and overall morphology of spin-coated perovskite films. Depending on the functionality of additives, uniform perovskite films can be obtained by providing homogeneous nucleation sites, and prolonged crystallization time can result in increased crystal size. [23,24,27] Inspired by these reports, we anticipate that the crystallization dynamics of perovskite films deposited by scalable methods can analogously be modulated by incorporating small amounts of additives into the precursor. This is the primary motivation that inspires us to formulate precursor inks by additive engineering. We choose MACl as the additive for our perovskite systems primarily based on the following reasons. First, MACl can effectively slow down the perovskite crystallization rate, which is expected to provide a viable approach to manipulate film morphology. [29] Second, the volatility of MACl allows the excessive additives to be evaporated from the perovskite film during thermal annealing. [30] Consequently, the phase purity and device performance will not be impaired by additives in the perovskite film. Significantly, we found that the incorporation of MACl plays an essential role in suppressing the formation of the yellow δ-phase during processing FACsinvolved perovskite films.
To begin with, we employ MAPbI 3 as our model material system in consideration of its simpler chemical composition compared to the multiple-cation and mixed-halide perovskites. It is thus expected to offer a facile platform in understanding the thin-film formation mechanism of the additive-and vacuum-assisted crystallization. We incorporated different levels of MACl, from 10% to 100% mole ratio relative to MAPbI 3 , into the precursor solution and deposited the precursor films by doctor-blade coating. The eventual crystalline perovskite films were obtained by vacuum extraction of the freshly deposited precursor films followed by a thermal-annealing step. Detailed procedures for perovskite film fabrication are presented in the Experimental Section and will be discussed in the next section.
We evaluate the robustness of our crystallization protocol by scrutinizing the microstructure, crystallinity, and optical properties of the polycrystalline MAPbI 3 films processed by blade coating. Figure 2a shows the top-view scanning electron microscopy (SEM) images of the perovskite films fabricated from precursor inks with different amounts of MACl. It is obvious that the crystal grain size enlarges substantially from ≈200 nm to over 4 µm with increasing content of MACl (Figure 2b), which confirms that the MACl additive plays an important role in nucleation and crystal growth. It has been demonstrated that perovskite ingredients dissolved in solution are mainly in the form of aggregated colloid particles rather than movable ions. [31] We therefore correlated the crystal size evolution to the coordination between the precursor solute and solvent, as manifested by the observation of Tyndall effect in the precursor solutions ( Figure S1, Supporting Information). It is also apparent that the additive has a significant effect on surface coverage of the films. Without MACl incorporation, several small holes with diameters of ≈200 nm are randomly distributed across the film. The existence of holes can deteriorate device performance by forming shunts. Remarkably, the incorporation of a small amount of 10% MACl produces uniform and dense perovskite films with complete coverage. Further increasing the MACl content to larger crystals but also introduces big uncovered areas, particularly for MACl loads of more than 50%. The formation of such voids presumedly results from the release of excess MACl during annealing processes. Crosssectional SEM images shown in Figure S2 of the Supporting Information evidence vertically aligned monolayer crystals with grain boundaries formed throughout the cross section for all levels of MACl content. In contrast, the perovskite film processed without additive shows randomly stacked small-size grains. We note that enlarged crystal grains with vertically stacked boundaries could benefit charge transport with minimal recombination at grain boundaries as the photogenerated carriers can travel through the active layer and reach the corresponding charge carrier extraction interfaces.
The structural and optical properties of the prepared perovskite films were inspected by X-ray diffraction (XRD) and ultraviolet-visible (UV-vis) absorption analyses with results presented in Figure 2c,d, respectively. We fixed the thickness of perovskite films at ≈450 nm by keeping all the deposition parameters constant. From the XRD patterns, we can see that the diffraction peaks at 14.1° and 28.2°, respectively, indexing to the (110) and (220) planes of MAPbI 3 are present for all perovskite films. The two diffraction peaks become much sharper and more intense with higher loads of additives, which is consistent with the enlarged grain sizes. For a better comparison, the XRD spectra were plotted with Y-axis in logarithm scale ( Figure S3, Supporting Information), where we can see that four additional diffraction peaks at 20.05° (112), 23.51° (211), 24.5° (202), and 31.92° (312) were observed for the perovskite without MACl additive. Upon addition of MACl, the intensity of the peaks slightly reduced in 10% MACl processed film which ultimately disappeared in the films with MACl loads higher than 20%. Particularly, the presence of the single two peaks at 2θ = 14.1° and 28.4° for the films with higher levels of MACl indicates a preferential crystal growth of (110) faces parallel to Crystallization is decoupled from precursor deposition Figure 1. The efficiency versus precursor deposition temperature for the preparation of perovskite solar cells via scalable blade coating or slot-die coating. It is noted that, although the work [1] (in the Supporting Information) reported the room-temperature deposition of perovskite precursor, the crystallization was carried out in an antisolvent bath which is not ideal for practical manufacture. [13] the substrate. Zoom-in spectra of the (110) and (220) planes shown in Figure S4 of the Supporting Information evidences negligible shift of the peaks, underlining that the addition of MACl does not significantly impact the MAPbI 3 crystal structure. Figure S5 of the Supporting Information displays the full width at half maximum (FWHM) of the (110) and (220) peaks of MAPbI 3 films as a function of MACl addition content, suggesting that MACl loads from 10% to 30% yield films with optimum crystallinity. From the UV-vis spectra shown in Figure 2d, we can see that the absorption of the perovskite films prepared with 10% to 30% MACl additives is slightly stronger than the pristine film. When MACl contents increased to 50% and higher, the absorption below 600 nm is strikingly reduced primarily due to the existence of large voids.
Considering that MACl has been previously reported to modulate the crystal morphology of the perovskite films based on spin-coating process, the direct transition of the precursor ink from spin-coating to our scalable printing methods demonstrated the unique advantages of our scalable deposition technology. This is indeed of vital importance for industrial production given that precursor inks developed for lab-scale spin-coating process can be directly transferred to scalable printing lines without redesigning the formulations. To further illustrate the generic application of the additive engineering in modulation of the crystal morphology of the printed perovskite films, we further prepared perovskite films with incorporation of additives of ammonium thiocyanate (NH 4 SCN) and potassium thiocyanate (KSCN) by blade-coating. [32] As presented in Figure S6 of the Supporting Information, the SEM images evidence that the addition of small amounts of NH 4 SCN and KSCN into the precursor solution can have analogous impacts on the film morphologies of blade-coated perovskite thin films.

Controlled Perovskite Crystallization
We now turn to illustrate the intractable yet of uttermost importance of the crystallization mechanisms for perovskite thin films prepared by blade coating.  crystallization protocol. The first step involves the deposition of a perovskite precursor film on the substrate via doctor-blade coating. The freshly deposited precursor film contains a large amount of solvent which requires to be judiciously removed as it would lead to unpredicted crystallization due to the simultaneous evaporation of solvent and solute upon thermal annealing. As mentioned earlier, this process is distinctly different to the previously reported approach to blade-coating at elevated temperatures, which are close to or exceeding the boiling point of solvents. [9][10][11][12][20][21][22] The crystallization process driven by such high temperatures will take place concurrently during precursor deposition. High temperature blading has the inherent disadvantage that film deposition is not decoupled from film drying and crystallization. With nucleation, precipitation, and mass transport happening simultaneously, process control becomes challenging and may result in undesired reproducibility or yield.
The classical antisolvent strategy provides a sound hint illustrating the importance of slowing down perovskite crystallization dynamics, which is realized by creating an intermediate phase upon pouring an immiscible solvent at a delayed spinning stage. [6] Enlightened by the antisolvent approach, we sought to achieve such an intermediate stage via vacuum extraction to remove excess solvent of the freshly deposited wet films. The resulting intermediate film could effectively separate the precursor film deposition from subsequent thermal annealing (Figure 3a), thus, allowing the precursor deposition to be carried out at room temperature. Finally, highly crystallized perovskite films with shiny surface were obtained by annealing the intermediate films at 100 °C for 10 min. We note that although vacuum extraction has been reported by several groups, it has been predominantly explored for preparation of perovskite films based on spincoating. [25,[33][34][35] In combination with additive engineering, Adv. Sci. 2019, 6,1901067  in the present work, we endeavor to exploit a generalized method that allows to precisely manipulate the morphology and crystallinity of perovskite films for a wide range of material compositions.
We highlight that the deployment of vacuum extraction and the incorporation of MACl are the two paramount important factors affecting the quality of the eventual perovskite films. As indicated in Figure S7 of the Supporting Information, thermally induced crystallization from the naturally dried perovskite precursors (without the vacuum extraction step) yield coarse surfaces with randomly aligned textured structure, regardless of MACl incorporation. In contrast, prior to thermal annealing and after a moderate vacuum process at 1000 Pa for 2 min, the resulting perovskite films with different dimensions (as large as 38.5 cm 2 ) show a mirror-like smooth surface and clear nonscattered transmitted light ( Figure S8, Supporting Information). Significantly, we observed an obvious color transformation from transparent yellow of the wet precursor film to shiny orange of the intermediate film during the vacuum process at around 1 min (Figure 3a). This observation was quite similar to the spin-coated perovskite films based on the antisolvent strategy. [6] Moreover, we found that the higher the level of MACl additive, the browner the intermediate films and the shorter time required for color change. We speculate that a certain degree of crystallization of perovskite took place in the intermediate films which was driven by the formation of supersaturation due to the vacuum removal of the excess solvent. To verify this hypothesis, we conducted SEM and XRD analyses to unveil the morphology and phase evolution of the intermediate films processed with 0%, 10%, and 30% MACl additives. SEM images shown in Figure 3b evidence that the intermediate film without additive is composed mainly of amorphous regions along with randomly aligned textured crystals. With incorporation of 10% MACl, noticeable cluster of grains appeared which are located on top of the textured crystals. Upon further increasing the MACl content to 30% cluster of grains become larger. The enlarged crystal gains in the intermediate films match well with the grain size evolution of the final crystalline perovskites. The formation of the perovskites was further confirmed by XRD measurement which is shown in Figure 3c. One can see that the characteristic perovskite diffraction peaks at 14.1° and 28.4° are present for all the three films and that the peaks become more intense and narrower with higher MACl content. These results attest that a mild vacuum at room temperature can induce the perovskite formation independent of additive incorporation, while the presence of MACl can promote the transformation of the precursor aggregates to crystalline perovskite possibly by reducing the activation energy for nucleation and crystallization.
In view of the importance of intermediate phases in regulating the morphology and crystallinity of the eventual perovskite films, we now take a close look at the structural evolution of the precursor films at different processing stages. We note that dimethylformamide (DMF):dimethyl sulfoxide (DMSO) with volume ratio of 4:1 was used as a solvent system for preparing our perovskite precursor inks, and it has been previously demonstrated that strong coordination between DMSO and PbI 2 via intercalation can benefit the perovskite film growth by retarding the reaction between the PbI 2 and MAI. [6,36] As indicated in Figure 3d, the XRD spectrum of the naturally dried precursor film (without vacuum extraction step) shows three diffraction peaks at low angles of 6.63°, 7.25°, and 9.23°. This crystalline complex can be ascribed to the intercalation of MAI and DMSO into layered PbI 2 with a molecular structure of MA 2 Pb 3 I 8 ·2DMSO, which has been resolved by several groups. [6,[37][38][39] Interestingly, upon a vacuum extraction process, these small angle peaks were obviously suppressed, which can be attributed to the collapse of some complex due to removal of DMSO. Simultaneously, two perovskite characteristic peaks (14.1° and 28.4°) emerged. It is also surprising to perceive that crystalline PbI 2 (diffraction 2θ = 12.7°) was negligible in both the intermediate film and the naturally dried precursor film. This observation implies that the PbI 2 in the intermediate film predominately forms crystalline MA 2 Pb 3 I 8 ·2DMSO complexes rather than bulk PbI 2 crystals. We further confirmed the presence of DMSO and MACl in the intermediate film with Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS) ( Figure S9, Supporting Information). When the intermediate film was subjected to an annealing step, highly crystalline perovskite was obtained, as evidenced by the disappearance of the low angle peaks with emergence of the two intense and sharp perovskite peaks (14.1° and 28.4°).
Having revealed the structural and morphological evolution of the printed films at different stages, the main reaction processes involved in our crystallization protocol can be proposed as follows (which is schematically illustrated in Figure 3e). i) The perovskite precursor film deposited by blade-coating contains excess solvent and the precursor ingredients in the wet film exist mainly in the form of aggregated colloid particles. ii) Once the wet precursor film is subjected to a vacuum extraction, the solvent DMF and excess DMSO are extracted from the wet film. The vacuum process creates a supersaturation stage, which not only transforms the precursor colloids to the intermediate film (MA 2 Pb 3 I 8 ·2DMSO), but simultaneously induces the formation of crystalline perovskite. iii) At last, applying a thermal annealing to the intermediate film, the MACl volatilized from the system and the residual DMSO was liberated from the MA 2 Pb 3 I 8 ·2DMSO complex and, simultaneously, PbI 2 reacted with MAI to produce tetragonal perovskite films.

Photovoltaic Performance at Different Device Scales
To evaluate the performance of the doctor-blade deposited MAPbI 3 films for photovoltaic applications, we fabricated solar cells with a device architecture of glass/ITO/PTAA/MAPbI 3 / PCBM/BCP/Ag (Figure 4a). The cross-sectional SEM image of a complete device with 10% MACl additive is shown in Figure 4b, which evidences micrometer-scale perovskite crystals monolithically aligned throughout the device. Figure 4c and Figure S10 of the Supporting Information present the statistic photovoltaic performance of the prepared MAPbI 3 solar cells with different amounts of MACl. Solar cells without MACl incorporation show rather low efficiencies of around 10%. Remarkably, 10% MACl mediated perovskites deliver the highest overall performance of around 17.5%, particularly in terms of high V OC of ≈1.00 V and fill factors (FF) up to 80%. However, when the MACl content exceed 20%, all the photo voltaic parameters tend to decline, which can be attributed to the presence of shunts and low light harvesting capacities due to the existence of large voids in the perovskites. Figure 4d shows the current density-voltage (J-V) curves of the best solar cell processed from 10% MACl, which gives an efficiency of 18.06% with J SC of 22.58 mA cm −2 , V OC of 1.00 V and FF of 80% obtained from the reverse scan. It is also noticeable that the 10% MACl mediated solar cells show no hysteresis which is in striking contrast with the solar cells processed without and with MACl contents higher than 20% ( Figure S11, Supporting Information). The high performance along with free of hysteresis of the device prepared from 10% MACl additive can be attributed to the high crystallinity and absence of pin-hole related recombination paths. [40] This can be evidenced from the SEM images (Figure 1a) that 10% MACl processed perovskite film showed complete coverage with compact crystal grains, while lots of pinholes appeared in the sample without MACl incorporation and the pinholes become larger with higher MACl additives. In Figure 4e, we show the external quantum efficiency (EQE) spectrum yielding an integrated J SC value of 21.85 mA cm −2 , which is in good agreement with the J-V measurement, with a discrepancy below 5%. We tested the steady state power output at maximum power point (MPP) of a blade-deposited solar cell and the result shows a small decline (current-density decreased from 20.7 to 20.2 mA cm −2 ) under continuous tracing under 1 sun illumination for 300 s (Figure 4f). To get a more comprehensive view on the stability of our solar cells, we performed a prolonged operational stability test of a device over a period of more than an hour. [41][42][43] As indicated in Figure S12 of the Supporting Information, the solar cell underwent a 6.3% degradation after 66 min continuous illumination, suggesting acceptable stability of our blade-deposited perovskite solar cells. Furthermore, the shelf stability of the devices stored in N 2 -filled glovebox was also evaluated. The result shown in Figure S13 of the Supporting Information indicates excellent shelf stability of the bladedeposited perovskite solar cells as almost no performance losses were observed over a month's storage.
To demonstrate the uniformity of the blade-coated perovskite films, we continue to produce solar cells with an active-area of 1 cm 2 (Figure 5a). Figure 5b shows the measured J-V curves which gives a PCE of 14.72% with V OC of 1.01 V, J SC of 21.70 mA cm −2 , FF of 67.15% (forward scan), and again the large-area device showed negligible hysteresis. The stabilized photocurrent of the device at the MPP (biased at 0.74 V) are shown in Figure 5c, yielding a stabilized PCE of 14.3%. The statistic device performance based on 15 solar cells from two batches is shown in Figure S13 of the Supporting Information, suggesting excellent reproducibility of the device fabrication and high quality of the blade-coated perovskite films. Compared with the small-size devices, the performance losses mainly come from the lower FF. [44] We therefore derived the series resistance (R S ) and parallel resistance (R P ) of the 1 cm 2 device from the measured J-V curves. As displayed in Figure S14 and Table S2 of the Supporting Information, the calculated R S is as high as 8.25 Ω cm 2 , which is more than four times larger than the small-size device, whereas the R P values for both devices are more than 1000 Ω cm 2 . These results indicate that, other than current leakage through shunts, the large series resistance mainly contributed to the lower FF of the 1 cm 2 -size solar cells, which was probably resulted from resistance losses due Adv. Sci. 2019, 6,1901067  to the high sheet resistance of ITO electrode (17 Ω sq −1 ). To confirm this hypothesis, we have built another 1 cm 2 solar cell on PTAA-coated ITO electrodes with a lower sheet resistance of 7 Ω sq −1 . As can be seen in Figure S14 of the Supporting Information, the device deposited on 7 Ω sq −1 ITO electrode shows a lower series resistance than that on 17 Ω sq −1 ITO. Accordingly, an enhanced FF of 72.6% was obtained (Table S2, Supporting Information).
Manufacturing of large-area modules is actively pursued on the way of realizing practical applications of thin-film photovoltaic technologies. The general challenge, however, is the quality and uniformity of the absorber films on large-scale levels. The highly reproducible fabrication of 1 cm 2 cells by blade-coating further encouraged us to manufacture modules with larger areas. We designed and fabricated 4 × 4 cm 2 modules with four cells monolithically interconnected in series (Figure 5d). The module has an active area of 10.08 cm 2 and an interconnection area of 2.88 cm 2 , leading to geometric fill factor of 77.78% ( Figure S15, Supporting Information). The J-V curves of the module in Figure 5e give a PCE of 11.25% (J SC = 4.93 mA cm −2 , V OC = 3.87 V, FF = 59%) from the reverse scan. However, a noticeable hysteresis was observed which is probably due to the inferior interconnection formation. We also tracked the PCE of the module near the MPP (biased at 2.85 V). The result indicates a slight current density increase during the illumination, leading to an efficiency of 11.01% under continuous light-soaking for 300 s (Figure 5f). It is worth noting that the voltage of the four-cell module reaches 3.87 V, suggesting negligible V OC losses (an average of 32.5 mV loss for the individual sub-cell) which again validates the high quality of the blade-coated perovskite film over the large area. The relatively low FF of the module may originate from the high resistance for charge recombination in the interconnection area where the P2 line was scribed manually with a tweezer.

Universality of the Crystallization Protocol
Having realized high-quality perovskite films using MAPbI 3 as a starting material, we now turn to illustrate the generality of the protocol for other perovskite material systems with particular interest in different band gaps. FAPbI 3 has a smaller band gap (1.50 eV) than MAPbI 3 (1.58 eV) and is thus beneficial for increased light harvesting. However, the transition from the α-phase to the unfavorable yellow δ-phase at low temperature renders the material not applicable for solar cells. Previous works have shown that addition of 5-20% Cs cations can suppress the formation of the δ-phase without significantly widening the band gap. [45,46] We therefore selected FA 0.95 Cs 0.05 PbI 3 as a smaller band gap material and tested the applicability of our vacuum-and additive-assisted crystallization strategy. Following the same procedure as for the preparation of MAPbI 3 films, we added different amounts of MACl into the FA 0.95 Cs 0.05 PbI 3 precursors to facilitate the formation of intermediate phases and examined the morphology and phase of the resulting perovskite films. As shown in Figure 6a, the SEM images reveal an overall crystal enlargement from ≈100 nm to ≈2 µm with increase of MACl additives. The XRD spectra shown in Figure 6b evidences increased characteristic diffraction peaks at 14.01° with higher loads of MACl, which is consistent with the enlarged crystal sizes. Surprisingly, it is found that, without addition of MACl, the presence of Cs alone is incapable of suppressing the formation of δ-phase (11.81°), as indicated by the coexistence of the yellow δ-phase and trigonal black FAPbI 3 α-phase (14.01°). This observation is strikingly different from the FA 0.95 Cs 0.05 PbI 3 films prepared by spin-coating coupled with antisolvent crystallization, where stable α-phase perovskite films are obtained without any additives. [46] Remarkably, with incorporation of 10% MACl, the δ-phase was largely suppressed, concurrent with the formation of a black α-phase perovskite. Further increasing MACl to 20%, the δ-phase was eliminated completely and pure α-phase FA 0.95 Cs 0.05 PbI 3 was obtained. The suppression of δ-phase of the FA 0.95 Cs 0.05 PbI 3 via incorporation of MACl can be presumably due to the formation of intermediate phase that favors the α-phase formation. To verify this hypothesis, we measured the XRD spectra of the intermediate films with different MACl levels. As shown in Figure S16  Overall, these results indeed underline the different crystallization mechanisms of the perovskite films produced by printing techniques compared to those of spin-coated films, which in turn highlights the necessity to re-engineer precursor inks as well as crystallization strategies for deposition of perovskite films by scalable methods. Figure 6c presents the UV-vis absorption of thin film FA 0.95 Cs 0.05 PbI 3 which clearly evidences a reduced band gap compared to MAPbI 3 . We prepared solar cells using perovskite films processed with 30% MACl additives to evaluate the photovoltaic performance of the blade-deposited cell extends to 850 nm, giving an integrated current density of 21.85 mA cm −2 ( Figure S17, Supporting Information). We continue to demonstrate the versatility of our method by fabricating a wide-band gap perovskite composite. Wider band gap perovskites are in the research focus for multijunction photo voltaics. Partial substitution of iodine with bromide can effectively broaden the band gap of the perovskites. We thus prepared mixed-halide perovskite films with a formula of FA 0.6 MA 0.4 Pb(I 0.6 Br 0.4 ) 3 , which has an optical band of 1.76 eV (Figure 6c and Figure S17, Supporting Information). As expected, SEM images ( Figure 6a) and XRD patterns of the prepared films ( Figure S18, Supporting Information) once again confirmed enhanced crystal size and crystallinity with addition of MACl. The best solar cell fabricated from 10% MACl additive yields a high V OC of 1.07 V, FF of 69.50%, and a J SC of 16.59 mA cm −2 , giving a PCE of 12.34% with a small hysteresis. The relatively low current density is mainly related to the narrow solar spectrum coverage due to its wide band gap ( Figure S17, Supporting Information).

Conclusion
In summary, we have developed a generic and robust crystallization technology specifically targeted at scalable deposition of high-quality perovskite films via printing methods. In contrast to previously proposed strategies with which the crystallization was directly driven by applying high temperatures during precursor deposition, the essence of our technology is the deployment of a vacuum-based precrystallization procedure. The vacuum extraction produces temporary intermediate films, allowing the precursor film deposition to be separated from the subsequent postannealing. Equally important, the incorporation of small amounts of MACl in precursor inks promotes the formation of intermediate films, which enables us to precisely control the morphology, phase, and crystallinity of perovskite films over a wide range of compositions. Multiple-cation and mixed halide perovskite composites with different band gaps have successfully been fabricated, attesting the general application of the technology for perovskite thin film processing. From the perspective of the global photovoltaics community, these encouraging results are anticipated to inspire the field to make a significant step toward the reproducible manufacturing of waferscale perovskite modules by high-throughput printing methods. Solar Cell Fabrication: The prepatterned indium tin oxide (ITO) coated glass (OPV Tech Co., Ltd.) was sequentially cleaned by sonicating the substrates in acetone and isopropanol for 10 min each. A PTAA holetransporting layer was spin-coated from 2.5 mg mL −1 CB solution on ITO substrate at 5000 rpm for 30 s. The film was then annealed at 120 °C for 5 min in ambient air. The substrate was transferred to a nitrogen-filled glovebox after it cooled down to room temperature. The perovskite absorber layer was subsequently deposited using the vacuum-assisted blade-coating method as described above. On top of perovskite film, the electrotransporting layer PC 61 BM (20 mg mL −1 in chlorobenzene) and the interfacial layer BCP (2.5 mg mL −1 in isopropanol) was successively deposited by spin coating at 2000 rpm for 30 s and 5000 rpm for 30 s, respectively. Finally, 100 nm Ag contact was deposited by thermal evaporation. The active areas of the solar cells were 0.09 and 1 cm 2 , respectively, for the small-size (MAPbI 3 , FA 0.95 Cs 0.05 PbI 3 , and FA 0.6 MA 0.4 Pb(I 0.6 Br 0.4 ) 3 ) and large-area (MAPbI 3 ) devices, which were determined by the overlapping between the top Ag and bottom ITO electrode. All the devices for performance and stability evaluation were tested without encapsulation.

Experimental Section
Large-area MAPbI 3 modules with four cells monolithic interconnected in series were fabricated on 4 × 4 cm 2 ITO-coated glass substrates. The P1 lines on ITO substrates with a space of 8 mm are prepatterned by laser scribing. On the cleaned ITO substrates, the hole-transporting layer PTAA was deposited by spin coating at 5000 rpm for 30 s. The perovskite film was then blade coated as detailed above. On top of perovskite layer, the electron-transporting layer PC 61 BM and buffer layer BCP were sequentially deposited by spin coating at 2500 rpm. for 35 s and 5000 rpm for 1 min, respectively. Three P2 lines were scribed mechanically using a tweezer parallel to the P1 lines. To complete the module fabrication, 100 nm Ag was thermally deposited using a shadow mask which was prepatterned to function the P3 lines. The series-interconnection of the module was realized by the mechanically separated but electrically interconnected P1, P2, and P3 lines ( Figure S15, Supporting Information).
Characterizations: Morphologies of the perovskite films were imaged with a scanning electron microscope (SEM, FEI Apreo LoVac). The structural properties of the perovskite films were analyzed by a Bruker D8 Advance X-ray diffractometer (XRD) using Cu Kα as the radiation source. The X-ray photoelectron spectroscopy (XPS) measurements were performed on an ESCALab250Xi electron spectrometer (Thermo Fisher) www.advancedscience.com using Al Kα radiation. Transmittance Infrared (IR) measurements were carried out employing an FTIR spectrometer (Thermo Scientific Nicolet 380). Optical absorption measurements were performed on a Agilent Cary 5000 spectrophotometer. The Tindall effect was tested with a commercial laser pen with green beam light. The current densityvoltage (J-V) characteristics and steady-state output of all the solar cells were measured using a Keithley 2400 source meter. The illumination was provided by a Newport Oriel 92 192 solar simulator with an AM1.5G filter, operating at 100 mW cm −2 , which was calibrated by a standard silicon solar cell from Newport. The EQE was taken using a QE-R instrument from Enlitech.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.