Author Correction: Efficient green light-emitting diodes based on quasi-two-dimensional composition and phase engineered perovskite with surface passivation

The original version of this Article omitted an acknowledgement to the source of Fig. 1a. The following has been added to the end of the caption to Fig. 1: 'Figure 1a adapted from ref. 13 (copyright 2016 Macmillan Publishers)'. This has been corrected in the PDF and HTML versions of the Article.

The exciton binding energy is enlarged in the two-dimensional (2D) perovskite and thus the emission efficiency can be enhanced 20,21 . In the early reports, 2D perovskite materials were used as emitting layers, while electroluminescence can only be observed at liquid nitrogen temperature 20 . Recently, quasi-twodimensional (quasi-2D) perovskites have been proposed by doping 2D perovskites into three-dimensional (3D) perovskite structures. The initial idea was to enhance the humidity stability of perovskite solar cells [22][23][24] . More recently, it was found that multiple quantum wells could be assembled in these quasi-2D perovskites 13,14 , which can facilitate the formation of exciton and reduce the possibility of exciton dissociation. This makes the quasi-2D perovskite an efficient luminescent material [13][14][15]25 . The working mechanism of quasi-2D PeLEDs has been fully discussed elsewhere 13,14 . Electrons and holes injected from large bandgap phases (small n-phase) can rapidly transfer into small bandgap phases (large n-phase), then electrons and holes recombine in large n-phase and emit as photons. The EQEs of quasi-2D perovskite LEDs have achieved around 10% 13,14 and highest EQE of 11.7% in near-infrared region was obtained 14 . Comparing to organic or inorganic LED counterparts 26,27 , there is still room for improvement in the EQE.
It is known that quasi-2D perovskite materials usually contain a mixture of phases and the phase impurity could lead to low emission efficiency, and thus fine control of the composition/ phases is critical for efficient emission. In addition, solution processed thin films of quasi-2D perovskite show reduced crystal size compared to those of 3D perovskite 13,15 . This could increase the concentration of the defects and traps on the film surface and grain boundaries, which will act as non-radiative recombination centers and thus decrease the emission efficiency. Surface passivation has been proved an efficient way for reducing the defects in perovskite solar cells [28][29][30] .

Results
Properties of quasi-2D perovskite with different compositions.
The schematics of n-phase PEA 2 (FAPbBr 3 ) n-1 PbBr 4 are drawn in Fig.1a, illustrating that the n-phase unit cell contains n-1 sheets of PbBr 6 -octahedra and two PEA tiers. The perovskite films are prepared by spin-coating a precursor solution using a crystalpinning process 11,32 . Their compositions are controlled by the ratio PEABr:FABr:PbBr 2 (Supplementary Table 1), and compositions n ranging from 2 to 6 are discussed. Previous results suggested that the quasi-2D perovskite films are not a single phase but contain a variety of n-phases 13,14 . For example, if the precursor solution with n = 2 composition is spin-coated, the generated films might include n = 2 phase but could also contain other phases such as n = 1 or n ≥ 3. To avoid misunderstanding, we distinguish the composition and phase by n-composition and n-phase, respectively. Methylammonium chloride (MACl) is also added into the precursor to improve the morphology and enhance the emission efficiency of the formed films (Supplementary Figs. 1, 2 and 12), but most MACl escape from the film during annealing 33 . The role of MACl has been well investigated in solar cells 33,34 . First, the MACl can slow down perovskite crystallization and improve film morphology. Second, the incorporation of a very small amount of elemental Cl could passivate the grain boundary, and then enhance the emission efficiency ( Supplementary Fig. 2). X-ray diffraction (XRD) measurements are carried out to determine the accurate phases in our obtained quasi-2D perovskite films. As shown in Fig. 1b, all the perovskite films with n ≥ 2 compositions show diffraction peaks at 14.8°, which is the same as the (100) diffraction patterns of the 3D perovskite FAPbBr 3 (refs. 7,35 ). However, the peaks become much more broadened, indicating that the diffraction peak could be from a series of PEA 2 (FAPbBr 3 ) n-1 PbBr 4 phases with large n (n→∞) or 3D FAPbBr 3 with nano-size grains ( Supplementary Fig.3) 15 . This result shows that all the quasi-2D perovskite films always contain large n-phases (n→∞) or 3D perovskite regardless of the composition of the precursor. Atomic force microscopy (AFM) images confirm the reduction of the crystal size during the transformation of 3D perovskite into quasi-2D perovskites ( Supplementary Fig.3).
In addition to the diffraction peak at 14.8°, a series of Bragg reflections at low angles (2θ < 14.8°) are observed for these quasi-2D perovskites (Fig.1b). This indicates that the PEA group with large size is incorporated, and the size of unit cells is enlarged compared with 3D perovskite 14,25,36 . According to the analysis shown in Supplementary Table 2 and the unit cells of n-phases shown in Fig.1a, it is found that the diffraction peaks (2θ < 14.8°) are attributed to the n = 1 and n = 2 phases (diffraction patterns from n ≥ 3 phases are not observed here). Specifically, for n = 1 composition, only the n = 1 phase exists; for n = 2 composition, the n = 1 and n = 2 phases are both presented in the films. It is unexpected that diffraction peaks of the n = 1 phase is absent in the n = 3 composition, while obvious diffraction peaks from the n = 2 phase are observed. For n ≥ 4 compositions (n > 4 compositions are not shown here due to very weak small n-phases diffractions), the diffraction peaks from small n-phases become weak or absent.
The photoluminescence (PL) results of the quasi-2D perovskite films with different compositions are shown in Fig. 1c, and the 2D (n = 1) and 3D (n = ∞) perovskite films are also included for comparison. The 2D perovskite shows an emission at 412 nm (3.01 eV) and the long tail 21,37 at a longer wavelength, which could be due to the self-trapped exciton emission 21 or the disordered structure. The 3D perovskite showed a band edge emission peak at 542 nm (2.29 eV) 38 . All quasi-2D perovskite films have two main emission peaks: one in the green region (approximately 532 nm, approximately 2.33 eV) and another in the blue region (at 440 nm, 2.82 eV). Comparing the PL with the absorption spectra ( Fig.1c and Supplementary Fig. 4), it can be found that the stronger green emission at approximately 532 nm comes from large n-phases (n→∞) or 3D perovskite, and the weaker blue emission at 440 nm comes from the n = 2 phase. The emission phases we find are nearly almost consistent with the XRD results. There are some exceptions: for the n = 2 composition, we have not observed the n = 1 phase emission, while the n = 4 composition shows an emission at 405 nm from the 2D perovskite phase (n = 1), and the reason for this is not clear at present.
The relative PL intensity and emission peak wavelength in the green emission from the quasi-2D perovskites and 3D perovskite are plotted in Fig. 1d. It is found that the n = 3 composition film shows the brightest green emission. The significant blue-shift of the green emission from quasi-2D perovskites is observed compared with the 3D perovskite. In addition, a smaller ncomposition showed a larger blue-shift, indicating that the blueshift is due to the quantum confinement effect. The emission images of these quasi-2D perovskite films under ultraviolet excitation (Fig. 1e) also confirms that the n = 3 composition shows the best PL.
Compared with 3D perovskites, significant enhancement of the PL is observed for quasi-2D perovskites (Fig.1c). This is due to the quantum confinement effect via the formation of quantum well structures between large bandgap phases (smaller n-phases) and small bandgap phases (n→∞ phase or 3D perovskite) 13,14 . The different emissions from various compositions observed in Fig.1c could be explained by the competition between confinement phase (small n-phase) and emission phase (large n-phase, n→∞ phase). Here, we use n = 2, 3, 4 compositions as examples. According to XRD results shown in Fig.1b, the n = 2 confinement phase and 3D emission phase dominate in n = 2 and n = 4 composition, respectively, and both of these will cause weak green emission. While the amount of n = 2 phase and 3D phase are adequate in the n = 3 composition, this could be the reason of the highest PL of n = 3 composition. The PLQY of the n = 3 shows a value as high as 57.3% under low density excitation (excitation wavelength of 400 nm, power density of 3.5 mW cm -2 ). The quantum well of n = 3 composition is mainly formed by the n = 2 and n→∞ phases ( Supplementary Fig. 6).
Passivation of quasi-2D perovskite. Keeping in mind that the n = 3 composition of PEA 2 (FAPbBr 3 ) n-1 PbBr 4 perovskite could be a good candidate for an emitting material, we further improve its emission by surface passivation. An organic small molecule TOPO is spin-coated onto the perovskite surface to form a thin passivation layer 31 . It is found that the PLQY of the n = 3 composition film was significantly increased from 57.3% to 73.8% after coating with the TOPO layer (Fig. 2a), confirming the passivation effect of the TOPO layer. The average fluorescence lifetime of the n = 3 composition film is also increased from 0.17 μs to 0.36 μs after TOPO treatment (Fig. 2b). Although a small red-shift in PL spectra has been observed for the TOPO-treated 3D perovskite film previously 31 , no obvious modulation of PL due to TOPO treatment is observed in this study ( Supplementary  Fig. 7).
The passivation of halide perovskites is ascribed to the chemical interaction between molecular ligands and incomplete PbI 6 -octahedra in previous study 39 . To confirm this, Fourier transform infrared (FTIR) spectroscopy measurements were carried out, and the result is shown in Fig. 2c. An absorption peak located at 1150 cm −1 is observed for TOPO, which corresponds to P=O bond stretching vibrations 40 . This P=O bond absorption peak shifts to approximately 1100 cm −1 in the film comprised of TOPO and PbBr 2 , indicating that the bonding between perovskite and TOPO could be formed 41,42 .  Fig. 3c. Although the perovskite films exhibit a mixture of many phases with different bandgaps, only single green EL is observed, even in logarithmic scale ( Supplementary Fig. 8), which is different from the PL. It can be estimated that the charge density injected by EL (10 17 s −1 cm −2 ) is two orders higher than that excited by PL (10 15 s −1 cm −2 ) under injection/excitation of 5.1 V, 30 mA cm −2 and 1 mW cm −2 , 340 nm excitation, respectively, while the blue emission is still not observed in EL. This could be due to the different working mechanism between EL and PL. For the PL, the photo-excited charges are driven from a larger bandgap phase to a smaller bandgap phase by their conduction/ valence band energy difference. However, for EL, the driving force is not only the energy difference, but also the applied external electric field. As a result, most charges will be injected and recombined in the smallest band gap phases. This could be the reason that only green EL is observed even under high charge density injection. The EL peak is located at 532 nm, and the full width at half maximum is about 23 nm. The emission of our PeLEDs exhibited a good color purity (95%) in the green region, with Commission Internationale de l'Eclairage (CIE) chromaticity coordinate at (0.21, 0.75) (Fig. 3d).
We characterize the PeLEDs with different compositions (Fig. 4a, b and Supplementary Fig. 9). It is found that the n = 2 composition showed the lowest injection current ( Supplementary  Fig. 9), which could be due to the poor transport properties of 2D phase. Correspondingly, the brightness and CE of n = 2 composition-based devices are both the lowest (Fig. 4a, b). The compositions of n = 3 and n = 4 based PeLEDs show the maximum brightness of about 7000 cd m −2 and 8700 cd m −2 , respectively, while n = 3 composition shows a higher CE (40.2 cd A −1 ) than n = 4 composition (37.6 cd A −1 ). For the n = 5 and n = 6 compositions, consistent with their PL efficiency (Fig.1c, e), their EL also could not compare with that of n = 3 composition. The detailed device performances for various compositions are summarized in Table 1. We next discuss the role of surface passivation in performance improvement for n = 3 composition. It is found that the turn-on voltage increases and the injection current decreases after introducing a TOPO layer (Supplementary Fig.10). This is due to the insulating property of TOPO and indicates that the TOPO is successfully coated on the perovskite surface. Figure 4c shows the brightness curve of the PeLEDs with and without TOPO. The maximum brightnesses of the PeLEDs with and without passivation are at the same level, while the devices with passivation showed higher CE (Fig. 4d). The best performing device without TOPO passivation shows a CE of 52.5 cd A −1 , while TOPO-treated devices show a CE as high as 62.4 cd A −1 , and the power efficiency also increased from 49.4 lm W −1 to 53.3 lm W −1 (Supplementary Fig.11). Accordingly, a high EQE of 14.36% is obtained (Fig. 4e). We also find that our devices show good reproducibility. The EQE histogram for 60 devices from five batches are presented, showing an average EQE of 13.1% with a low relative standard deviation of 4.4% (Fig. 4f).

Discussion
Device stability is still a critical and common issue in PeLEDs. The stability of our PeLEDs with encapsulation was tested under a constant injection current in dry air (Supplementary Fig. 14). The luminance and CE increased at the beginning for several minutes, which could be due to trap filling 15 that results in the improvement of electrical contacts. After that, the device performance gradually decreased, and finally the devices work for approximately 120 min. We also noticed that larger injection current induces faster degradation, which could be due to heat accumulation or ion movements. Our device showed the almost same level stability of I-related perovskite with the stability in several 100 min 15 . Compared with the Br-related perovskite with several 10 min stability in previous reports 15,44 , our devices showed a little bit improvement.
Further improvement of PeLED stability is necessary, and we assume that there are several issues leading to PeLED instability. One major issue could be ion migration under voltage bias, and another issue could be the phase stability of perovskite materials under heating or moist conditions. Several strategies could be adopted to improve PeLED stability such as finding stable perovskite-emitting materials 19,45 , incorporating suitable groups to avoid or suppress the ion movements 12,19 , seeking effective ligands 17,46 to bond with the grain surface and also the film surface to reduce Joule heating induced by non-radiative recombination and finding stable charge injection layers 47 .
In summary, by composition/phase engineering, we find that a quasi-2D perovskite of PEA 2 (FAPbBr 3 ) n-1 PbBr 4 with n = 3 composition showed good PL. We also introduce surface passivation to reduce the non-radiative recombination on the perovskite surface or grain boundaries. Combining composition/phase engineering and surface passivation, high performance of quasi-2D PeLEDs with a CE of 62.4 cd A -1 and an external quantum efficiency of 14.36% is achieved. Device fabrication. The indium tin oxide (ITO)-coated glass substrates were sequentially cleaned in detergent, distilled water, acetone and isopropanol by sonication and used as anode. The cleaned substrates were ultraviolet ozone treated for 15 min to make the surface hydrophilic, then m-PEDOT:PSS (or normal PEDOT:PSS) aqueous solution was spin-coated at 9000 rpm (or 2500 rpm) for 40 s and baked at 160°C for 15 min in ambient air. Thereafter, the substrates were transferred into a nitrogen-filled glove box, and the 0.6 mol l −1 (Pb 2+ concentration) perovskite solution in DMSO was spin-coated onto the m-PEDOT:PSS (or normal PEDOT:PSS) films at 3000 rpm for 2 min, and after spin coating for 40 s, 100 µl of CB was poured onto the film for pinning the perovskite crystallization, followed by annealing on a hot plate at 90°C for 1 h. The perovskite precursor solution was prepared according to Supplementary Table 1. For surface passivation, 2 mg ml −1 TOPO in CB was dripped onto the perovskite film for a 1 min to let TOPO passivate the film surface enough and then spin-coated at 7000 rpm for 1 min. Finally, the fabrication of PeLEDs was completed by depositing TPBi, LiF (less than 1 nm) and Al electrode layer by layer through a shadow mask in a high vacuum thermal evaporator. The device area was 0.108 cm 2 as defined by the overlapping area of the ITO and Al electrode. The thickness of ITO and Al are 130 nm and 60 nm, respectively.
Material and device characterizations. XRD measurements were performed with a Rigaku D/max 2500H equipment with a conventional Cu target X-ray tube (Cu  K-alpha, λ = 1.5418 Å) set to 40 kV and 200 mA. Steady-state PL spectra and timeresolved fluorescence spectra (TRPL) of the perovskite films were measured at room temperature in the ambient air using a fluorescent spectrophotometer (HORIBA JY Nanolog-TCSPC). The excitation wavelength for PL and TRPL are both 340 nm, which were provided by 450 W xenon lamp and a pulsed diode laser (NanoLed-340), respectively. The PLQY of the perovskite films was measured using an Edinburgh FLS920 equipment with a 450 W xenon lamp as excitation source. The following settings were applied for PLQY measurements: excitation wavelength of 400 nm; bandpass values of 3.00 and 0.35 nm for the excitation and emission slits, respectively; step increments of 0.5 nm and integration time of 0.3 s per data point. We measured the light intensity with a laser power meter and the estimated light power density incident on the samples was about 3.5 mW cm −2 . The ultraviolet-visible absorbance spectra were recorded on an ultraviolet-visible spectrophotometer (Cary 5000, Varian). The FTIR spectra were recorded on an infrared spectrophotometer (Excalibur 3100, Varian) in a general mode that did not use attenuated total reflection component. AFM images of the perovskite films were collected in noncontact mode (Bruker FASTSCANBIO). The cross-sectional image of PeLEDs was made on a scanning electron/focused ion beam double-beam equipment (FEI Nova200 NanoLab) operated at 2 kV. Ultraviolet photoelectron spectroscopy (UPS) spectra were collected on a Thermo Scientific ESCALab250Xi equipment with an applied bias of −10 V. He I ultraviolet radiation source (21.22 eV) was used. The helium pressure in the analysis chamber during measurement was about 2E−8 mbar.
The film samples over the ITO layer had a conductive connection with an Au sample, so the Fermi level value of the film samples is equal to that of Au sample. The work function ϕ (that is the Fermi level absolute value of the free film) of the test films can be calculated from following equation: hν − ϕ = E Fermi -E cutoff , where E Fermi and E cutoff are respectively the value of Fermi level position and the steep edge position in the UPS spectrum of the test film, hν = 21.22 eV, E Fermi is 21.02 eV for Au. Two Keithley 2400 source meter units linked to a calibrated silicon photodiode were used to measure the current-voltage-brightness characteristics. The measurement system was carefully calibrated by a Hamamatsu C9920-02G equipment and a 100 mm integrating sphere of Enli Technology using our perovskite LEDs. Lambertian profile was used in the calculation of EQE. All the device characterization tests of perovskite LEDs were carried out at room temperature in the ambient air for the un-encapsulated devices, except that operational stability test was carried in dry air for the encapsulated devices.
Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request.