a Schematics of formation of the perovskite layer. b Cross-section STEM image of an ITO/ZnO/perovskite/Au device (scale bar, 100 nm) and the corresponding EDS characterization from the areas marked in STEM image (scale bar, 50 nm), and EDS profiles of Pb, Br, Zn, and C elements derived from the vertical white lines indicated on EDS mapping image. c AFM height images and multiple line scans of the FAPbBr₃ films. Scale bar, 500 nm. ATR-FTIR spectra (d) and UV-Vis absorption spectra (e) of the perovskite films at different annealing times, and the absorption spectra of 10 nm ZnO layers and DMSO-rinsed perovskite films (inset in (e)).

We then proceed to monitor the crystallization process of the 3D FAPbBr₃ perovskite films by using a range of film characterization techniques. Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectra (Fig. 1d) of the perovskite films showed that the amide signal become evident and remains almost the same after 1 min of annealing. Notably, the absence of these peaks in the control sample without ZnO underscores that the sacrificial ZnO layer is essential for initiating the interfacial amidation reaction (Supplementary Fig. 3a). Ultraviolet-visible (UV-vis) absorption spectra revealed that the interfacial reaction leads to the formation of 3D FAPbBr₃ perovskites and the consumption of the ZnO substrates. The absorption spectra (Fig. 1e) of the perovskite films showed that the strong absorption peaks at higher energies gradually disappear while the perovskite absorption edge at around 532 nm becomes dominant within the annealing time of 1 min, suggesting the transformation of low-dimensional phases into 3D perovskite films^16,17,26. We then washed off the perovskite films with a dimethyl sulfoxide (DMSO) solution to expose the underlying ZnO layers. For the perovskite films annealed for 1 min, the corresponding DMSO-rinsed sample showed an insignificant absorption signal (inset in Fig. 1e), indicating that the ZnO layer was completely consumed during the interfacial amidation reaction. The photoluminescence (PL) spectra (Supplementary Fig. 4a) showed that the PL peak exhibited a slight redshift upon increasing the annealing time of the perovskite films. X-ray diffraction (XRD) patterns (Supplementary Fig. 4b) confirmed the existence of low-dimensional perovskite phases at the initial annealing times. In addition, the perovskite films exhibited significant enhancement in crystallinity with the increased annealing time, as evidenced by the rapid increase of the XRD peak intensity. We note that XRD patterns of control samples without ZnO showed no appearance of the perovskite phase (Supplementary Fig. 3b), highlighting the critical role of the interfacial amidation reaction on the formation of 3D FAPbBr₃ perovskite. AFM images (Supplementary Fig. 5) further revealed that as the annealing time increased, the perovskite grains grew from smaller sizes into larger and more faceted structures. We notice once again that the Zn_x(Amide)_yBr_z complex remains around the perovskite grains during their formation and growth, passivating the grains like other Zn-amine complex²⁷.

To evaluate the emission properties of the FAPbBr₃ perovskite films, we first carried out time-dependent PL measurements on the films under continuous annealing at 100 °C. By tracking the PL intensities over time (Fig. 2a), we observed that the samples maintained ~90% of their initial PL intensity after 60 min of annealing, indicating their decent thermal stability at elevated temperatures. Their thermal stability was further confirmed by their almost unchanged absorption properties, crystallinity, and morphological characteristics after annealing at 100 °C for 60 min (Supplementary Fig. 6). PLQY measurements showed that FAPbBr₃ films displayed a high PLQY reaching 80% (Fig. 2b). We also observed a trap saturation process characterized by a notable increase in PLQY at extremely low excitation intensities in a range of ∽0.1-1 pJ cm⁻² (Fig. 2b), corresponding to ∽10¹⁰-10¹¹cm⁻³ excited carrier density (Supplementary Note 2 and Supplementary Fig. 7). Importantly, the PLQY of the films can be maintained at a high value of ~70% across a broad excitation range from 1 pJ cm⁻² to 0.2 μJ cm⁻² (Fig. 2b), suggesting that the trap-assisted nonradiative recombination is strongly diminished in 3D FAPbBr₃ films. Power-dependent time-resolved photoluminescence (TR-PL) characterizations further validated their low trap density. The PL decay collected at the lowest carrier density of 2.7 × 10¹¹cm⁻³ showed a negligible trapping behavior, indicated by the absence of sharp PL decay at the initial time (Inset in Fig. 2c). Importantly, the PL decay time of these FAPbBr₃ films was noticeably longer compared to other reports^26,28,29, showing a value of 4.5 μs within the linear PLQY range (Supplementary Note 3 and Supplementary Table 2). TR-PL measurements at different excitation intensities (Fig. 2c) indicated that the Auger recombination starts to dominate at an excited carrier density of 5.5 × 10¹⁵cm⁻³. By fitting the power-dependent integrated PL data using a generic model³⁰, the trap density of 3D FAPbBr₃ films was determined to be 1.2 × 10¹⁰cm⁻³ (Supplementary Note 4 and Supplementary Fig. 8), which is comparable to the value reported for FAPbBr₃ single crystals (9.6 × 10⁹cm⁻³)³¹.

Fig. 2: Luminescence properties of 3D FAPbBr₃ films.

a Thermal PL stability of FAPbBr₃ films under continuous annealing at 100 °C. Insets show PL spectra of the perovskite films at two different annealing times and a photograph of the green emitting films under 365 nm UV excitation. b Excitation-intensity-dependent PLQYs of the perovskite film. c Time-resolved PL decay transients of the perovskite film under different excitation intensities. The inset shows the initial PL decay collected at the lowest carrier density of 2.7 × 10¹¹cm⁻³.

We further noticed that the key to depositing high-quality FAPbBr₃ films on ZnO substrates lies in precisely controlling the thickness of the oxide layer, which ensures efficient interfacial amidation and complete consumption of the sacrificial ZnO layer. To illustrate this fact, we deposited 9 wt.% perovskite precursor solution on top of the 5 nm and 20 nm ZnO layers, respectively. For the samples deposited on 5 nm ZnO substrates, both the absorption spectra and XRD pattern (Supplementary Fig. 9) revealed the existence of a significant portion of low-dimensional perovskite phases. In contrast, the perovskite films deposited on 20 nm ZnO substrates showed pure 3D perovskite features (Supplementary Fig. 10). However, these films had reduced thermal stability, as evidenced by a dramatic decrease in PL intensity under continuous annealing at 100 °C (Supplementary Fig. 11). Additionally, upon removing the perovskite film deposited on 20 nm ZnO through rinsing with DMSO, ~10 nm of unreacted ZnO was revealed (Supplementary Fig. 12). This residual ZnO may cause continuous deprotonation of the FA⁺ cations and the subsequent decomposition of the perovskite films¹⁶.

All-inorganic architecture perovskite light-emitting diodes

The decent PL properties and the significantly improved stability of FAPbBr₃ films prompted us to use them in perovskite LEDs. We have proposed a device architecture featuring a perovskite emitting layer sandwiched between a 2-(3,6-dibromo-9H-carbazol-9-yl) ethyl] phosphonic acid (Br-2PACz) SAM and a bilayer oxide EIL comprising 10 nm ZnO-S and 10 nm ALD-ZnO. The Br-2PACz modified ITO substrate, with a work function of 5.5 eV, ensured efficient hole injection (Supplementary Fig. 13). Cross-sectional STEM images of these LEDs display a distinct island-like structure of perovskite (Fig. 3a), which is consistent with observations from the ITO/ZnO/perovskite/Au device (Fig. 1c) and is beneficial for enhancing light out-coupling efficiency⁴. We note that the sacrificial ZnO layer was completely consumed during the interfacial reaction, and thus it does not serve as a charge-transporting layer in the operational device.

Fig. 3: All-inorganic architecture of perovskite LEDs based on FAPbBr₃ films.

a Cross-sectional STEM image of a full all-inorganic architecture device. Scale bar, 100 nm. b J-V curves for the electron-only devices (ITO/Al/oxide EILs/Ag) based on the 50 nm ZnO-S films and bilayer 10 nm ZnO-S/40 nm ALD-ZnO films. c PLQYs for the pristine FAPbBr₃ film, the FAPbBr₃/ALD-ZnO film, the FAPbBr₃/ZnO-S film, and the FAPbBr₃/ZnO-S/ALD-ZnO film deposited on quartz substrates. Error bars correspond to the standard deviation. d EL spectra of a perovskite LED under different luminance. Inset shows a photograph of an operating green-emitting device with the City University of Hong Kong (CITYU) logo. e J-V-L curves of the champion device. f EQE, and PCE versus the current density characteristics of the champion device. g Stability data at various initial brightness. The inset shows acceleration factors (n) fitted based on the empirical relationship ?0??50 = constant.

We further designed a bilayer ZnO-S/ALD-ZnO EIL to minimize interfacial quenching while ensuring efficient electron transport. Previous studies found that ZnO-S EILs exhibited minimal quenching of the perovskite emission but suffered from low electrical conductivity due to the long carbon chain ligands on the ZnO-S surface²⁵. To address this issue, we incorporated an ALD-ZnO layer to enhance electron transport. Electrical analyses of electron-only devices with an ITO/Al/oxide EILs (~50 nm)/Ag structure demonstrated a significant (two orders of magnitude) increase in current densities upon introducing the ALD-ZnO layer (Fig. 3b). PLQY measurements of the FAPbBr₃ films in contact with different oxide EILs showed that direct deposition of an ALD-ZnO layer on the FAPbBr₃ film resulted in severe interfacial emission quenching, leading to rather low PLQY of 39% (Fig. 3c). SEM characterizations revealed that the interfacial quenching originated from the damage to perovskite layers during the ALD process, which was significantly mitigated by introducing a ZnO-S interlayer (Supplementary Fig. 14). As a result, the FAPbBr₃ films interfaced with the bilayer ZnO-S/ALD-ZnO EIL maintained a high PLQY of 71% (Fig. 3c), suggesting significantly suppressed interfacial quenching at the FAPbBr₃/bilayer ZnO interface.

The perovskite LEDs employing the ZnO-S/ALD-ZnO EIL exhibited remarkable optoelectronic properties. They maintained identical green EL peak at 540 nm with a narrow full width at half maximum (FWHM) of ~21 nm across a wide brightness range from 100 to 300,000 cd m⁻² (Fig. 3d), showcasing the high color purity and distinguished spectral stability. The angle-dependent EL intensity followed a Lambertian profile (Supplementary Fig. 15). Figure 3e presents the current density-voltage-luminance (J-V-L) characteristics of these LEDs, featuring a sub-bandgap turn-on voltage of only 1.9 V (Fig. 3e). The current density increased rapidly once the devices were turned on, yielding a maximum luminance of up to ~312,000 cd m⁻². Additionally, due to the suppression of interfacial luminance quenching and efficient electron transport, these devices achieved low driving voltages of 2.3 V and 2.7 V for brightness levels of 10,000 and 100,000 cd m⁻², respectively (Fig. 3e), both of which are superior to state-of-the-art green perovskite LEDs (Supplementary Table 3).

As a result of low operation voltages, these perovskite LEDs exhibited a peak PCE of 18.2% and maintained a high PCE of 15.6% at a high current density of 300 mA cm⁻² (Fig. 3f), suggesting low-efficiency droop. It is noteworthy that this performance is significantly better than in previous reports, where the highest PCEs achieved for green-emitting perovskite, organic, and quantum dot LEDs without out-coupling enhancement at 300 mA cm⁻² were only 13.5%⁶, 2.7%³², and 5.5%²³, respectively. Additionally, our devices achieved a peak EQE of 20.7% (Fig. 3f), and the analysis of 50 devices fabricated using the same procedures revealed an average EQE of 17.9% with a standard deviation of 2.6% (Supplementary Fig. 16), confirming decent reproducibility.

Importantly, these perovskite LEDs not only exhibited high PCEs but also demonstrated significantly extended operational lifetimes. The T₅₀ lifetime is plotted in Fig. 3g for devices operated at different initial brightness levels (L₀). Notably, the T₅₀ lifetime was 350 h at an initial brightness of 1000 cd m⁻² (Fig. 3g). By fitting the T₅₀ data with an empirical equation ((L₀)ⁿ × T₅₀ = constant) across a broad range of L₀ from ~1000 to ~20,000 cd m⁻², we determined an acceleration factor (n) of 1.69. Thus, the projected T₅₀ operational lifetime for the perovskite LEDs was estimated to be 17,000 h at 100 cd m⁻².

For comparative analysis, a control perovskite LED without the sacrificial ZnO layer was fabricated, which failed to work due to the unsuccessful formation of the perovskite active layer (Supplementary Fig. 17). Additionally, control devices employing either ZnO-S or the ALD-ZnO films as EILs were also fabricated. Perovskite LEDs with ALD-ZnO EILs failed to operate due to the destruction of the perovskite layer during the ALD process, while devices incorporating ZnO-S EILs exhibited current densities significantly lower (by one order of magnitude) than those with bilayer ZnO EILs (Supplementary Fig. 18a). The decreased current densities resulted in a significant EQE roll-off (a PCE of ~5% at 80 mA cm⁻²) and a markedly reduced T₅₀ lifetime of 25 h at 1000 cd m⁻² (Supplementary Fig. 18c,d). These results underscore the importance of the interfacial reaction and the necessity of using bilayer ZnO EILs to achieve high-performance perovskite LEDs.

Potential of fabricating perovskite films for lasing

To achieve high-quality and thicker FAPbBr₃ films for lasing applications, it is crucial to manage the concentration of perovskite precursors and the thickness of the ZnO substrate. As mentioned above, using a 20 nm ZnO substrate with a 9 wt.% precursor resulted in ~10 nm of unreacted ZnO. To address this issue, we increased the precursor concentration to 18 wt.%. Figure 4a displays the absorption spectra of the 18 wt.% perovskite films before and after rinsing with DMSO. The absorption spectra of the pristine perovskite films confirm the formation of 3D FAPbBr₃ (Fig. 4a), which is further verified by their XRD patterns (Supplementary Fig. 19). After DMSO rinsing, the absorption spectra showed negligible absorption from the ZnO layer (inset in Fig. 4a), suggesting the complete consumption of the sacrificial ZnO substrate. Similar to the thinner perovskite films demonstrated above, the 18 wt.% perovskite films also exhibited decent stability, retaining over 90% of their initial PL intensity after continuous annealing at 100 °C for 60 min (Fig. 4b). Notably, these perovskite films also showed a high PLQY of 77% and a low trap density of 1.7 × 10¹⁰cm⁻³ (Supplementary Fig. 20). Benefiting from the remarkable optical properties, the devices based on 18 wt.% perovskite films achieved a peak EQE of 19.4% (Supplementary Fig. 21).

Fig. 4: Characterization of the 18 wt.% FAPbBr₃ film deposited on the 20 nm ZnO substrate.

a UV-vis absorption spectrum of the 18 wt.% FAPbBr₃ film. The inset shows the spectrum of the corresponding DMSO rinsed FAPbBr₃ film. b PL spectra of perovskite films under continuous annealing at 100 °C at two different times. c Evolution of the PL spectra of the 18 wt.% FAPbBr₃ film under increasing pump fluence of the femtosecond laser excitation (with a central wavelength of 400 nm and 1 kHz repetition rate). d PL intensities versus the pump fluence, and the determined ASE threshold.

Encouraged by the achieved low trap density of the FAPbBr₃ films, we proceeded to study their potential lasing properties. Figure 4c shows the PL spectra of the 18 wt.% perovskite films at various pump fluences. At an excitation fluence of 4.2 μJ cm⁻², the emission spectrum shows a conventional PL profile with a peak at 544 nm and a FWHM of ~24 nm (Fig. 4c). Upon increasing the pump fluence, a sharp and intense narrow PL peak emerges at ~555 nm and becomes more prominent, while its FWHM decreases sharply to 8 nm, indicative of amplified spontaneous emission (ASE). The ASE threshold for the 18 wt.% perovskite films was determined to be 13 μJ cm⁻² (Fig. 4d), which is comparable to the state-of-the-art ASE thresholds for cavity-free perovskites emitting in the green part of the spectrum (Supplementary Table 4).

In summary, we introduced a method to significantly enhance the performance of green perovskite LEDs through the optimization of interfacial chemistry and device architecture. By employing a straightforward interfacial amidation reaction on sacrificial ZnO layers, we successfully fabricated high-quality FAPbBr₃ perovskite films with a controllable thickness. These films exhibited a low trap density (~10¹⁰ cm⁻³), comparable to that of single FAPbBr₃ perovskite crystals. We integrated these FAPbBr₃ films into a perovskite LED with a mono-molecular HIL and an all-inorganic bilayer EIL configuration. The resultant devices not only achieved champion values for brightness (312,000 cd m⁻²) and PCEs (15.6% at 300 mA cm⁻²) but also set benchmarks for perovskite LEDs in terms of extended operational stability (T₅₀ lifetime of 17,000 h at 100 cd m⁻²). Moreover, the optimized perovskite films demonstrated a low ASE threshold of 13 μJ cm⁻², which opens up avenues for the development of perovskite-based electrically pumped lasers, further broadening the practical applications of perovskites in optoelectronic devices.