a, Photograph of a perovskite film on a 2 cm × 2 cm glass/ZnO-PEIE substrate, alongside a coin. The perovskite film is shiny and uniform.b, Optical microscope images with different magnifications. The scale bars represent 30 μm.c, SEM images with different magnifications. The scale bars represent 3 μm.d, High-magnification SEM images of randomly selected regions. The scale bars represent 3 μm.e, AFM images with different magnifications. The scale bars represent 2 μm. The images show the submicrometre-scale structure of the perovskite film.

a, SEM images of perovskite films. The scale bars represent 1 μm. The images show that as the annealing time increases, the crystallites grow from small particles and become larger and more faceted. When the annealing time is more than 6 min, similar submicrometre-scale structures form.b, XRD spectra. Crystallinity is enhanced as the annealing time increases, in agreement with the SEM images.c, Excitation-intensity-dependent PLQEs. Trap densities gradually decrease as the annealing time increases, resulting in PLQEs of more than 60% when the annealing time is between 16 and 20 min.d, Time-resolved photoluminescence (PL) decay transients (at a carrier density of 1.0 × 10¹³ cm⁻³). The films show longer PL decay lifetimes at longer annealing times.e, Dependence of current density and radiance on the driving voltage. The circles denote a bunch of curves; the arrows show they axis to which a given bunch belongs.f, EQE versus current density.g, Peak EQE versus annealing time. An average peak EQE of more than 18% can be maintained with annealing times of between 14 and 20 min. Error bars correspond to the standard deviation.

a, Dehydration reaction of 5AVA on top of the ZnO-PEIE surface^36,37.b, Grazing-angle reflectance FTIR spectra of perovskite films at various annealing times. As the annealing time increases, the peak at 3,400–3,300 cm⁻¹ (ν_O–H in 5AVA) decreases; simultaneously, peaks appear at 1,620 cm⁻¹ and 1,550 cm⁻¹, which can be assigned as amide I band (ν_C = O) and amide II band (δ_N–H), respectively. These spectra indicate dehydration reactions of 5AVA, leading to the formation of an organic layer during annealing.

The ratio of 5AVA to FAI to PbI₂ isx/2.4/1, wherex varies from 0 to 0.9.a, SEM images. The scale bars represent 1 μm. The value ofx is given in the top left corner of each image. The reference FAPbI₃ perovskite film without 5AVA has low film coverage. Without 5AVA, the perovskites form discrete clusters with random shapes. After adding 5AVA, faceted perovskites with submicrometre structures gradually form.b, XRD spectra. The perovskite films show improved crystallinity with the addition of 5AVA.c, Excitation-intensity-dependent PLQE. After adding 5AVA, PLQEs were greatly enhanced, indicating reduced trap densities.d, Time-resolved PL decay transients (carrier density 1.0 × 10¹³ cm⁻³). There is a fast PL decay channel for the perovskite without 5AVA, indicating a high level of trap densities. This fast PL decay channel gradually disappears after adding 5AVA.e, Dependence of current density and radiance on the driving voltage. After adding 5AVA, the leakage current is reduced.f, EQE plotted against current density.g, Peak EQE plotted against 5AVA ratio. Error bars correspond to the standard deviation. After adding 5AVA, the peak EQE increases owing to reduced leakage current and enhanced PLQE. When the 5AVA ratio is increased to 0.9, the EQE decreases, owing to the inferior outcoupling efficiency that results from the more dispersed structural pattern.

a, Device structure. A typical reference device consists of a metal layer (Au), a 7-nm-thick MoO₃ layer, a 40-nm-thick TFB layer, a 50-nm-thick emitting layer (EML), a 30-nm-thick layer of ZnO-PEIE, a 160-nm-thick ITO layer and a semi-infinite glass substrate. In our new device, the EML is replaced by a layer of perovskite squares distributed with a periodP and a duty cyclel_e/P (wherel_e is the length of the perovskite platelets, andl_e/P = 50%). The height of the convex structure of TFB is denoted ash and the diameter is set tol_e + 100 nm.b, Discretized map of the perovskite layer. The scale bar represents 1 μm.x andy are the pixel numbers in units of pixel lengtha.f(x,y) is the discrete function.c, Module of spatial frequency spectrum.U_x andU_y are the spatial frequencies.d, Refractive indices of different layers in our perovskite LEDs. Optical constants (n,k) of the multilayers were determined using an ellipsometer. Here the optical constants of perovskite are from a continuous FAPbI₃ film, which are used in the simulation.e, EQE calculated as the periodP and the convex heighth.f, Calculated outcoupling efficiency as a function of periodP with convex heighth = 30 nm. The reference is a device made from continuous perovskite film. The simulation shows that the outcoupling efficiency can be more than 25% over a wide range of periods from 310 nm to 900 nm.

Measuring the device EQE at low temperatures minimizes nonradiative recombination so that the EQE reaches a value of 30% at 6 K.

a, SEM images. The scale bars represent 1 μm. As the precursor concentration increases (shown in the top left corner of each image), the size of the crystallites increases and the crystallites become more tightly packed.b, XRD spectra.c, Excitation-intensity-dependent PLQEs. The 7 wt.% film shows the highest PLQEs.d, Dependence of current density on driving voltage. The leakage currents of the devices decrease as the precursor solution becomes more concentrated.e, Dependence of radiance on the driving voltage. As the precursor concentration increases, the turn-on voltage increases and the radiance decreases, probably owing to the poor charge transport of thicker film.f, EQE plotted against current density.g, Peak EQE versus precursor concentration. Error bars correspond to the standard deviation. When the concentration exceeds 10 wt.% the EQE decreases, probably because of a reduced outcoupling-enhancement effect and poor charge transport.

a, 5 wt.%.b, 7 wt.%.c, 10 wt.%.d, 15 wt.%.e, 20 wt.%. Charge carrier densities vary as indicated (black, blue, purple and green traces).

a, SEM image of submicrometre-structured perovskites fabricated with 6ACA (chemical structure shown in white). The scale bar represents 1 μm. Inset, FFT pattern in a randomly selected region. TheP range of 6ACA is 265–901 nm, yielding a calculated outcoupling efficiency of 28.9% ± 2.5%.b, SEM image of submicrometre-structured perovskites fabricated with 7AHA (chemical structure shown in white). The scale bar represents 1 μm. Inset, FFT pattern in a randomly selected region. TheP range of 7AHA is 432–1,430 nm, yielding a calculated outcoupling efficiency of 26.4% ± 3.3%.c, Excitation-intensity-dependent PLQE. The perovskite films with 6ACA and 7AHA have similar PLQEs.d, Dependence of current density and radiance on the driving voltage.e, EQE versus current density. The 6ACA- and 7AHA-based devices reach peak EQEs of 18.2% and 17.3%, respectively. Given that the perovskite films based on 6ACA and 7AHA have similar PLQEs, the EQEs must be affected mainly by the different outcoupling efficiencies that result from the different periodicities of the submicrometre-scale structures.f, Electroluminescence spectra.