Large-area and efficient perovskite light-emitting diodes via low-temperature blade-coating

Large-area light-emitting diodes (LEDs) fabricated by mass-production techniques are needed for low-cost flat-panel lighting. Nevertheless, it is still challenging to fabricate efficient large-area LEDs using organic small molecules (OLEDs), quantum dots (QLEDs), polymers (PLEDs), and recently-developed hybrid perovskites (PeLEDs) due to difficulties controlling film uniformity. To that end, we report sol-gel engineering of low-temperature blade-coated methylammonium lead iodide (MAPbI3) perovskite films. The precipitation, gelation, aging, and phase transformation stages are dramatically shortened by using a diluted, organoammonium-excessed precursor, resulting in ultra-flat large-area films (54 cm2) with roughness reaching 1 nm. The external quantum efficiency of doctor-bladed PeLEDs reaches 16.1%, higher than that of best-performing blade-coated OLEDs, QLEDs, and PLEDs. Furthermore, benefitting from the throughput of the blade-coating process and cheap materials, the expected cost of the emissive layer is projected to be as low as 0.02 cents per cm2, emphasizing its application potential. Fabrication of efficient large-area perovskite light emitting diodes catering towards mass-production is hampered by the difficulty in getting homogenous uniform films and the high cost. Here, the authors demonstrate sol-gel engineering of low-temperature blade-coated perovskite films which can overcome these limitations.

Metal halide perovskites (MHPs) have become a new generation of light-emitting materials due to their fantastic optoelectronic properties, such as high color purity, tunable band gaps, bipolar conduction, etc. [11][12][13][14][15] . The EQE of perovskite LEDs (PeLEDs) has been boosted to >20% with a small device area of <0.1 cm 2 (refs. [16][17][18][19][20][21][22][23][24][25] ) and has recently reached 12.1% with an area of 9 cm 2 using spin-coating 26 . Nevertheless, perovskite films made by mass-fabrication techniques such as blade-coating are usually very rough and not continuous due to coffee ring effect and fast crystallization of perovskites 27 . Blade-coating at high temperatures above 100°C, incorporation of additives/surfactant, or assistance of air knife have been developed to improve the morphology of thick perovskite films for solar cells 28,29 . However, the thick films with large grains are not suitable for PeLEDs due to their low electron-hole capture rates 20 . Therefore, the EQE of PeLEDs made by mass-production techniques reached only 1.1%, much lower than the devices fabricated by spin-coating 30 .
In this work, we report highly efficient PeLEDs fabricated by a robust blade-coating approach at a low temperature of 50°C. The sol-gel stages of the doctor-bladed methylammonium lead iodide (CH 3 NH 3 PbI 3 , MAPbI 3 ) films are effectively modified by changing precursor concentration and incorporation of excess bulky organoammonium halides. The diluted precursor with excess organoammonium halides results in much denser nucleation centers, slower/eliminated gelation process, and faster phase transformation process. As a result, the roughness of perovskite films decreases dramatically from 394.7 to 8.7 nm with precursor concentration reducing from 0.8 to 0.02 M and further decreases to around 1 nm with 50% molar excess of bulky organoammonium halides. The large-area, blade-coated films show great uniformity in terms of thickness, roughness, and optoelectronic properties. After incorporating the uniform film as an emissive layer, the EQE of PeLEDs reaches 16.1% with an area of 0.04 cm 2 and 12.7% with an area of 1 cm 2 . The PeLEDs with an ultra-large area of 28 cm 2 with uniform emission is also demonstrated. Furthermore, the cost of the emissive layer reaches 0.02 cents per cm 2 , demonstrating its huge potential in real applications of flat panel lighting and display.

Results
Doctor-blading of ultra-thin uniform perovskite films. Figure 1a shows a schematic of blade-coating process. The depth of applicator is fixed around 2-3 µm to ensure relatively thin perovskite films. The concentration of MAPbI 3 precursors varies broadly from 0.02 to 0.8 M in dimethylformamide (DMF). A N 2 knife is applied to accelerate solvent evaporation and improve film morphology. No antisolvent is used in the blade-coating process. It should be noted that the blade-coating speed, temperature, and the N 2 knife pressure affect the film morphology dramatically ( Supplementary Figs. 1-3). An ultra-fast coating speed of 150 mm/s, corresponding to Landau-Levich mode that the as-coated layer is still wet right after blading 29 , a low temperature of 50°C, and a N 2 knife pressure of 0.2 MPa are adopted for the coating process. Details of the optimization process can be found in Supplementary Information. As shown in Fig. 1b, the thicknesses of the films linearly depend on the precursor concentration, resulting from the Landau-Levich coating mode, and are slightly thinner with N 2 knife. Incorporation of bulky organoammonium halides, 4-fluorophenylmethylammonium iodide (FPMAI), does not change the thickness obviously. The thicknesses of the blade-coated films are very repeatable with acceptable fluctuations ( Supplementary Fig. 1).
Atomic force microscopy (AFM) and scanning electron microscopy (SEM) measurements are performed to examine surface morphology of the blade-coated films. It is unexpected that the perovskite grains become smaller and smaller, and the films become more and more uniform with precursor concentrations decreasing from 0.8 to 0.02 M ( Fig. 1c and Supplementary  Fig. 4a). Accordingly, the film roughness reduces dramatically from 394.7 nm (film thickness~950 nm) to 8.7 nm (film thickness 25 nm). The films prepared with the aid of N 2 knife are smoother with smaller grains due to the faster DMF evaporation speed and higher nucleus density ( Fig. 1d and Supplementary  Fig. 4b). The roughness reduces accordingly from 60.8 to 1.6 nm with precursor concentrations decreasing from 0.8 to 0.02 M. Nevertheless, the rod-like grains for the thick films suggest that N 2 knife will not affect grain growth mechanism.
Interestingly, after incorporation of 50% molar excess FPMAI, the perovskite grains change from rod-like to circle-like (Fig. 1e, Supplementary Fig. 5a), clearly indicating that the grain growth of perovskite is suppressed. No low-dimensional perovskite X-ray diffraction (XRD) peaks are observed in the film with FPMAI ( Supplementary Fig. 7), indicating that FPMA ligands work as surfactant self-assembled at the three-dimensional perovskite nanocrystal surface. Notably, the film roughness is dramatically reduced from 131.8 to 4.3 nm with precursor concentration decreasing from 0.8 to 0.02 M. The roughness is further reduced to 0.8 nm using a 0.02 M precursor solution with N 2 knife ( Fig. 1f and Supplementary Fig. 5b). A high-resolution SEM image shows that the grain sizes are around 10 nm ( Supplementary Fig. 6), comparable to the optimized spin-coated films 20 . The roughness of the films prepared by different methods and precursor concentrations are plotted in Fig. 1b for comparison.
Sol-gel stages of the film formation process. In order to figure out the origin of improved film morphology using very dilute precursor solutions and with excess FPMAI, we examined the film-forming process using an optical microscope (Supplementary Movies 1-4). The snapshots of the videos are shown in Supplementary Fig. 8. We draw schematics of the film formation process using the sol-gel knowledge according to the videos (Fig. 2). Four stages of sol-gel process can be clearly observed from the videos for the dense precursor (0.8 M) without excess FPMAI or N 2 knife. Stage 1. PbI 2 precipitation. The as-coated wet film/ink-layer quickly becomes saturated for PbI 2 due to its relatively lower solubility than the organic component. Therefore, PbI 2 precipitate first in the form of PbI 2 -DMF-MAI complex 31 . This stage starts at around 4.3 s after blade-coating with nucleus density around 1.43 × 10 2 /mm 2 ( Supplementary Fig. 8a). Stage 2. Gelation. The PbI 2 -DMF-MAI nucleus grows quickly because of the saturated PbI 2 wet film/ink-layer. Stage 3. Aging. With DMF continuing to evaporate, the PbI 2 -DMF-MAI complex continues growing and small PbI 2 -DMF-MAI complex grains start to precipitate. In addition, the PbI 2 -DMF-MAI complex starts to convert to perovskite phase as the concentration of organic component starts increasing. Stage 4. Phase transformation. As most DMF evaporate out, PbI 2 continues reacting with MAI and forms perovskite. This step finishes at around 14.7 s.
Excitingly, the sol-gel stages change substantially for the dilute precursor (0.2 M) (Supplementary Movie 2). The schematics of the film-forming process are shown in Fig. 2b. Stage 1. PbI 2 precipitation. With DMF evaporating, PbI 2 precipitates near the surface. But PbI 2 precipitations do not grow as the wet film/inklayer is not PbI 2 saturated. This stage starts at around 4.4 s. Stage 2. More PbI 2 precipitation and gelation. With DMF continuing evaporating, more PbI 2 precipitates near the surface and the existing PbI 2 precipitates grow slowly ( Fig. 2b and Supplementary  Fig. 8b). As a result, the nucleus density reaches 1.85 × 10 4 /mm 2 , which is two orders magnitude higher than the case using dense precursor ( Supplementary Fig. 8b). Stage 3. Aging. The PbI 2 -DMF-MAI complexes continue growing slowly and start to convert to perovskite phase. Stage 4. Phase transformation.

ARTICLE
Notably, the duration of this stage is very short because of the high nucleus density. This stage finishes at around 6.8 s. More excitingly, the sol-gel stages are totally different after incorporation of excess FPMAI in the precursors (Supplementary Movies 3 and 4). The schematics of film formation process using a dilute precursor (0.2 M) are shown in Fig. 2c. Stage 1. PbI 2 precipitation/phase transformation. The PbI 2 precipitation and phase transformation process happen simultaneously at the surface, indicated by the black nucleus at the beginning ( Supplementary Fig. 8c, d). This is because the excess organoammoniums in the precursor promotes reaction between PbI 2 and MAI/FPMAI. It also should be noted that the timing for PbI 2 precipitation is 7.5 s, in contrast to 4.4 s for the case without FPMAI. This is because excess FPMAI can increase the solubility of PbI 2 . Stage 2. More PbI 2 precipitation and phase transformation. With DMF evaporating, more PbI 2 (or MAPbI 3 grain) precipitates near the surface. It should also be noted that the existing MAPbI 3 grains do not obviously grow in this stage because bulky FPMA ligands suppress grain growth. This results in a high nucleus density of 4.53 × 10 5 /mm 2 ( Supplementary  Fig. 5a), more than one order magnitude higher than the case without FPMAI. Stage 3. Aging and phase transformation. As DMF fully evaporate out, the perovskite film with small grains forms. N 2 knife accelerates DMF evaporate speed and induce much denser nucleus. Therefore, the film formation processes are dramatically shortened and cannot be resolved by in situ optical microscope.
Large-area perovskite films fabricated by doctor-blading. To manifest the robustness of our blade-coating process, large films are prepared with an area of 6 × 9 cm 2 using the 0.04 M precursor solution. As a comparison, we also prepared films with the same size and thickness using a solvent-exchange approach, one of the most reliable methods to prepare high-quality uniform films. Notably, a higher precursor concentration of 0.2 M should be used to fabricate films with similar thickness with blade-coating (~40 nm). Furthermore, only a small volume of precursor (30 µl) is needed for blade-coating, while around 300 µl is required for spin-coating to ensure that the precursor can spread over the whole substrate. The applicator depth is critical to fabricate largearea uniform films. The 10 µm applicator results in precursor flowing on the substrate and non-uniform film ( Supplementary  Fig. 9). As shown in Fig. 3a, b, the blade-coated film is very uniform, while the spin-coated film shows circle-like morphology, possibly resulting from different solvent evaporation speed between the center and edge area during the spinning process.
The macroscale and microscale morphology uniformity of the large-area films are examined by thickness profiler and AFM, respectively. The large-area films were cut into small pieces with 1 × 1 cm 2 in size and marked as A xy (x, y = 1, 2, 3…) for the convenience of characterizations (Fig. 3c). Notably, the thickness of the bladed-coated film is very uniform over the whole substrate, while the spin-coated film is thinner at the center and thicker at the edge area (Fig. 3d, e). Excitingly, the bladecoated film is extremely uniform even at the microscale over the  Fig. 5 and Movies) of the film formation processes without N 2 knife. The film formation process with N 2 knife or using dilute solutions with concentration <0.2 M are too fast to be resolved by optical microscope. The timing for precipitation beginning and phase transformation ending are also marked below the schematics.
whole substrate with roughness around 1 nm (Fig. 3f). In contrast, the roughness of the spin-coated film reaches 4-5 nm at the center and >10 nm at the edge area (Fig. 3g).
We then examine the uniformity of optoelectronic properties of the large-area films. Figure 4a, b show photoluminescence (PL) images of blade-coated and spin-coated films under an ultraviolet lamp. The PL emission of the blade-coated film is very uniform, agreeing well with its uniform morphology. In contrast, the PL emission is weaker in the edge area for the spin-coated film. This should be due to its rough morphology and/or insufficient DMF extraction during the solvent-exchange process. We then measured PL quantum yield (PLQY) and time-resolved PL (TRPL) of the films. The steady-state PL spectra and TRPL decay curves are shown in Supplementary Fig. 10. Excitingly, the bladecoated film is very uniform in steady-state PL peak positions and intensities, as well as in TRPL decay. In contrast, the spin-coated films show relatively large fluctuations. The PLQY and average carrier lifetime (T average ) mapping of the large-area films are shown in Fig. 4c-f. The blade-coated film shows higher PLQY and longer T average than the spin-coated film with better uniformity. The result of microscale PL intensity mapping at the center area of the large films also demonstrates that the bladecoated film has stronger PL emission than the spin-coated film ( Supplementary Fig. 11). Figure 4g, h and Supplementary Tables 1  and 2 show radiative and non-radiative recombination rates calculated using equation 1 The blade-coated film shows relatively lower nonradiative recombination and higher radiative recombination than the spin-coated film, which is critical for light-emitting devices.   The device structure is shown in Fig. 5a. The molar ratio of FPMAI is fixed at 50% as it results in higher EQE than other ratios ( Supplementary Fig. 12). As shown in Fig. 5b, all PeLEDs with FPMAI and N 2 knife have very low leaking current in the range of 10 −5 mA/cm 2 , resulting from the uniform and compact film. Notably, the highest EQE reaches 16.1% using 0.04 M precursor (Fig. 5c), higher than that of the state-of-the-art spincoated devices based on MAPbI 3 (ref. 32 ). The electroluminescence (EL) spectra of the PeLEDs are shown in Fig. 5d. The blue shift of EL peaks for the ultra-thin perovskite films using dilute precursors should result from dielectric confinement effect by the bulky ligands self-assembled at the perovskite nanocrystal surface and/or adjacent organic electron/hole transporting layers 33 , considering that the grain sizes are larger than the Bohr radius 34 . The angular intensity profiles and EL spectra are shown in Supplementary Fig. 13. The EL spectra of PeLEDs with different thicknesses overlap for different angles. The EQE statistics of PeLEDs fabricated by different methods are shown in Fig. 5e and Supplementary Fig. 16. The devices without additive show relatively poor performances with EQE <1% ( Supplementary  Fig. 14a-d). This should be due to the rough films as well as low electron-hole binding energy of large MAPbI 3 grains. The EQE are generally higher with thinner films, resulting from more uniform film and better confinement of the electrons and holes. Spin-coating Blading direction 9 cm Fig. 4 Optoelectronic characterizations of large-area perovskite films. a, b PL images of the large-area perovskite films fabricated by doctor-blading (a) and spin-coating (b). The film area is 6 × 9 cm 2 . c-h PLQY (c, d), T average (e, f), and radiative and non-radiative recombination rate (g, h) mapping of largearea films fabricated by doctor-blading (c, e, g) and spin-coating (d, f, h). The excitation wavelength and intensity for PLQY (TRPL) are 430 nm (369 nm) and 8 mW/cm 2 (6 mW/cm 2 ), respectively. The characterizations are performed on small-area films (56 films in total). The three-dimensional figures are plotted by spatial interpolation. range of 1-10% even without N 2 knife for different perovskite thickness ( Supplementary Fig. 14e, f), resulting from their relatively small grains.
Notably, the performance of our doctor-bladed PeLEDs has exceeded that of best-performing doctor-bladed OLEDs and PLEDs, which have been intensively studied for decades 8,9 . Figure 5f compares the peak reported EQEs of various types of LEDs fabricated by blade-coating. The superior performance of blade-coated PeLEDs should result from the extreme uniform blade-coated perovskite films fabricated by sophisticated control of sol-gel processes. The sol-gel engineering approach is expected to be applicable to other types of solution-processed LEDs, including OLEDs, QEDs, and PLEDs, to improve their performances as they are generally also sol-gel processes.
To further demonstrate the robustness of doctor-blading of PeLEDs, large-area devices are fabricated on both rigid glass/ITO substrates and flexible polyethylene naphthalate (PEN)/ITO substrates. The layout of our large-area devices is shown in Supplementary Fig. 15a. The poly-TPD layer is also fabricated by doctor-blading with a thickness of around 30 nm to ensure largearea uniformity. As shown in Supplementary Fig. 15b, the doctorbladed poly-TPD layer is very flat with roughness around 0.2 nm, comparable to the spin-coated films. Notably, the devices show a decent EQE of 12.7 and 7.1% on rigid and flexible substrates, respectively (Fig. 5e, f). The insets show photo images of the large-area PeLEDs with a device area of 1 × 1 cm 2 on glass/ITO substrates and flexible PET/ITO, respectively. An ultra-large PeLEDs with a device area of 4 × 7 cm 2 is also fabricated. As shown in Fig. 5i, the device has uniform EL emission over the whole working area, demonstrating the possibility of large-area production of PeLEDs. We finally measured the operational lifetime of the PeLEDs with different areas. As shown in Supplementary Fig. 17, the operational lifetime reached 19.6, 11.0, and 10.1 h for device area of 0.04, 1, and 28 cm 2 , respectively, comparable to the state-of-the-art spin-coated devices 26 . Furthermore, the doctor-bladed PeLEDs also has superior advantages in terms of cost, which is critical for flat panel lighting applications. The cost of the emissive layer is calculated to be around 0.02 cents per cm 2 , demonstrating huge potential applications of PeLEDs in flat panel lighting.

Discussion
In conclusion, we have shown that the sol-gel stages of the bladecoating process of perovskite films can be effectively modified by changing precursor concentration and incorporation of excess  precursors before blade-coating in air at 50°C. The humidity during blade-coating was around 35%. The speed of applicator and N 2 knife was set as 150 and 50 mm/s, respectively. The pressure of N 2 knife was 0.2 MPa. The distance between the N 2 knife and substrate is 2 cm. After blade-coating, the perovskite films were transferred to N 2 -filled glovebox immediately. For doctor-blading of poly-TPD layer, the poly-TPD solution (6 mg/ml in chlorobenzene) was doctor-bladed using the same method with perovskite film. The amount of poly-TPD and perovskite solutions were about 0.5 μl/cm 2 .
Spin-coating of perovskite films. In all, 50% molar ratio excess FPMAI were added into MAPbI 3 solutions (0.2 M in DMF) before spin-coating. The spin coating rate was 4000 rpm, and toluene was dropped on the spinning substrate at around 6 s. About 300 μl of MAPbI 3 solution was used to prepare large-area (6 × 9 cm 2 ) films.
Characterizations of perovskite films. The thickness of perovskite films was measured using a Dektak XT profiler. The AFM measurements were conducted with tapping mode (Asylum Research MFP-3D-SA). The SEM measurements were conducted using field-emission SEM (Hitachi SU4800). XRD measurements were performed with a TTR-III (Rigaku) X-ray diffractometer with Bragg-Brentano parallel beam geometry, a diffracted beam monochromator, and a conventional Cu target X-ray tube set to 40 kV and 200 mA. The PLQY of the perovskite films were measured using a Horiba Fluorolog-3 system with Petite Integrating Sphere. The excitation wavelength was 430 nm with an excitation intensity of 8 mW/cm 2 . The TRPLs were taken using a Horiba timecorrelated single-photon counting system. The samples were excited by a pulsed laser diode (NanoLED-Horiba) with a center wavelength of 369 nm, an excitation intensity of~6 mW/cm 2 , and a repetition rate that is less than the reciprocal of the measurement range. The time resolution was determined to be~100 ps from the instrument response function. All decays were well fitted by 3 exponentials: decay t ð Þ ¼ P 3 n¼1 B n e Àt=T n , T average ¼ The PL intensity was recorded by a hybrid PMT detector, and the PL mapping was conducted over a 10 × 10 μm region.
PeLED fabrication and characterization. Poly-TPD was dissolved in chlorobenzene at a concentration of 6 mg/ml. The Poly-TPD solution was spin-coated on ITO substrate at 1000 rpm, followed by annealing at 150°C for 20 min. The poly-TPD layer was treated by O 2 plasma for 6 s to improve wetting. All bladecoated perovskite films are dried at 70°C for 10 min to fully dry the film and ensure full reaction of the precursors without affecting the morphology and grain size substantially. The TPBi, LiF, and Al layers were sequentially thermally evaporated on top of the perovskite film with thicknesses of 40, 1.2, and 100 nm, respectively. The device areas were 0.04, 1, and 28 cm 2 , respectively. The PeLEDs were measured in N 2 using a homemade motorized goniometer set-up consisting of a Keithley 2400 sourcemeter unit, a calibrated Si photodiode (FDS-100-CAL, Thorlabs), a picoammeter (4140B, Agilent), and a calibrated fiber optic spectrophotometer (UVN-SR, StellarNet Inc.). The distance between LED device and photodetector is 59.5 mm.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.