Downscaling micro- and nano-perovskite LEDs Yaxiao Lian1,#, Yaxin Wang1,#, Yucai Yuan1,#, Zhixiang Ren1,#, Weidong Tang1, Zhe Liu1, Shiyu Xing1, Kangyu Ji2, Bo Yuan1, Yichen Yang1, Yuxiang Gao1, Shiang Zhang1, Ke Zhou1, Gan Zhang1, Samuel D. Stranks2,3, Baodan Zhao1,* & Dawei Di1,* 1. State Key Laboratory of Extreme Photonics and Instrumentation, College of Optical Science and Engineering, ZJU-Hangzhou Global Scientific and Technological Innovation Center, International Research Center for Advanced Photonics, Zhejiang University, Hangzhou, China. 2. Cavendish Laboratory, University of Cambridge, Cambridge, UK. 3. Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, UK. # These authors contributed equally to this work. *Corresponding authors, E-mail: daweidi@zju.edu.cn (D.D.); baodanzhao@zju.edu.cn (B.Z.). Many technological breakthroughs in electronics and photonics were made possible by downscaling – the process of making elementary devices smaller in size1-5. The downsizing of light-emitting diodes (LEDs) based on III-V semiconductors led to micro-LEDs5-12, an ‘ultimate technology’ for displays. However, micro-LEDs are costly to produce and they exhibit severe efficiency losses when the pixel sizes reduce to ~10 μm or below, hindering their potential in commercial applications. Here, we show the downscaling of an emerging class of LEDs based on perovskite semiconductors to below the conventional size limits. Micro- and nano-perovskite LEDs (micro/nano-PeLEDs) with characteristic pixel lengths from hundreds of μm down to ~90 nm are demonstrated, through a localised contact fabrication scheme that prevents non- radiative losses at the pixel boundaries. For our near-infrared and green micro-PeLEDs, average EQEs are maintained at around 20% across a wide range of pixel lengths (650 to 3.5 μm), exhibiting minimum performance reduction upon downsizing. Our nano-PeLEDs with characteristic pixel lengths of down to ~90 nm represent the smallest LEDs reported, enabling a record-high pixel density of 127,000 pixels per inch (PPI) amongst all classes of LED arrays. Our demonstration showcases the strength of micro/nano-PeLEDs as a next-generation light source technology with unprecedented compactness and scalability. 2 Downscaling of electronic devices is an everlasting pursuit in information science and technology, leading to revolutions in computational power and human-machine interactions1-5. Micro light- emitting diodes (micro-LEDs)5-9, evolved from conventional LEDs based on III-V semiconductors10- 12 through downsizing to the microscale, exhibit enhanced light-emitting performance featuring high brightness, high resolution, low energy consumption, and rapid response13-19. These advantages empowered the recent development of near-eye display technologies including virtual reality (VR) and augmented reality (AR)20. The production processes of micro-LEDs involve epitaxial growth, top- down etching and mass transfer. The difficulties in the reliable processing of micro-LEDs intensify with downscaling, leading to limited yields and substantially increased production costs5,8. Micro- LEDs suffer from significant efficiency losses when the pixel sizes reduce to near or below 10 μm5,8,21- 23. Important progresses were made in achieving high device performance at reduced lengthscales through the preparation of nanostructured LEDs based on III-V materials24,25. However, the highly demanding fabrication processes for these LEDs limit the long-term prospects for commercial applications. As an emerging LED technology, perovskite light-emitting diodes (PeLEDs) are known to exhibit high performance at low processing costs26-30. Since the demonstration of room-temperature electroluminescence (EL) from halide perovskites in 201426, substantial advances have been made in device performance. High external quantum efficiencies (EQEs) of 20-30%31-35 and operational lifetimes approaching those of organic LEDs were demonstrated36-38. The concept of micro- PeLEDs39,40 with pixel dimensions on the order of 100 µm or below, was proposed as a new direction of PeLED research. This type of device shows potential to be an alternative to conventional micro- LED technology at greatly reduced fabrication costs. Initial prototypes of micro-PeLEDs were demonstrated using self-aligned photolithography39, giving peak EQEs of ~4.1% for a pixel area of 200´100 µm2. More recently, PeLEDs with a pixel area of 20´5 µm2 were achieved by direct optical patterning41,42, yielding a peak EQE of 6.8% and a maximum luminance over 20,000 cd m-2. Furthermore, micro-PeLEDs with a pixel area of 78.5 µm2 were achieved by stamp patterning43, yielding a peak EQE of 3.9% and a maximum luminance of 400 cd m-2. Similar to micro-LEDs, when the pixel sizes of micro-PeLEDs reduce to tens of μm or below, significant efficiency losses limit the possibilities of further downsizing. In this work, we report the downscaling of micro- and nano-perovskite light-emitting diodes (micro/nano-PeLEDs) with characteristic pixel lengths (Lch, defined by Lch=Area1/2 of each pixel) from 3 hundreds of µm down to ~90 nm. The devices are lithographically patterned to feature localised electrical contacts that define the active pixels and prevent non-radiative losses at the pixel boundaries. For our near-infrared (NIR) and green devices, average EQEs are maintained at around 20% across a wide range of characteristic lengths (650 to 3.5 μm), exhibiting minimum performance reduction upon downsizing. The characteristic lengths for the average EQEs to reduce to 50% of their maximum values (L50) are found to be ~180 nm and ~440 nm for NIR and green devices, respectively. Nano-PeLEDs with characteristic pixel lengths of down to ~90 nm are achieved, representing the smallest LEDs reported to date. This translates to a pixel density of 127,000 pixels per inch (PPI), a record amongst all classes of LED arrays. A prototypical active-matrix micro-PeLED display based on a commercial TFT array is developed. Our work demonstrates the exceptional potential of micro- and nano-PeLEDs for next-generation display technologies. Results and discussion The efficient micro/nano-PeLEDs are enabled by the fabrication processes illustrated in Fig. 1a (see Methods for details). The device architecture of the micro/nano-PeLEDs consists of glass/ITO/SiO2/ electron-transport layer (or hole-transport layer)/perovskite/hole-transport layer (or electron-transport layer)/metal electrode (Fig. 1b). Micro/nano-patterning processes are involved in the preparation of three functional layers, including the ITO substrate, the SiO2 insulating frame and the metal electrode. The lithographically patterned SiO2 (iii-viii in Fig. 1a) serves the key purpose of defining active pixels by forming localised contacts that are away from the edges of the electrodes, ensuring the patterning accuracy and edge regularity of the active pixels (Fig. 1c,d and Extended Data Fig. 1a-b). This device fabrication process is referred to as the “localised contact” scheme. Without the lithographically defined SiO2, the irregular edges of the ITO electrodes (Fig. 1e,f) severely limit the patterning accuracy of the devices. In contrast, the edges of the pixel areas produced by the localised contact scheme are clearly defined (Extended Data Fig. 1c,d and Extended Data Fig. 2a-c). High-quality micro/nano pixels with characteristic lengths of down to ~90 nm can be obtained (Fig. 1g,h and Extended Data Fig. 1e,f). The sidewall/pixel area ratios for different pixel sizes are calculated. The ratios are low (<5%) for pixel sizes ranging from 650 μm to 1 μm (Extended Data Fig. 2f). When the pixel length reduces to ~100 nm, the sidewall/pixel area ratio reaches 20% (Fig. 1h(iii)), approaching the limit of reliable fabrication using the method we developed. 4 Fig. 1 | Fabrication processes of micro- and nano-PeLEDs. a, Key fabrication steps of micro- and nano-PeLEDs. (i) ITO patterning by laser etching. The width of each ITO stripe is 150 μm. (ii) The SiO2 layer (~130 nm) was deposited using plasma enhanced chemical vapor deposition (PECVD). (iii) Positive photoresist was deposited by spin-coating. (iv) For micro-PeLEDs, the microscale pixel areas were exposed by ultraviolet light; for nano-PeLEDs, the nanoscale pixel areas were patterned by dual-beam focused ion beam (FIB) etching method. (v) Development of positive photoresist (for micro-PeLEDs). (vi) Etching of the SiO2 layer by reactive ion etching (RIE). (vii) Removal of the residual photoresist. (viii) The surface was passivated using a thin SiO2 layer (~5 nm) by magnetron sputtering. (ix) Sequential deposition of bottom CTL, perovskite, and top CTL. (x) Electrode deposition through a metal mask. b, Schematic cross-sectional view of the micro- and nano-PeLED. c-d, AFM images of the patterned electrodes with and without using the localised contact method. e-f, Height profiles corresponding to the lines in (c) and (d). g, SEM images of (i) square and (ii) circular pixel areas after RIE and photoresist removal. h, SEM images of the localised contact regions of circular pixels with characteristic lengths of 890, 440 and 90 nm, corresponding to diameters of 1000, 500 and 100 nm, respectively. The uniformity of the perovskite near the pixel edges defined by the SiO2 frame (prepared by the localised contact method) is found to be superior to those prepared without using this approach (Extended Data Fig. 2d,e). We note that the perovskite formed inside the active pixel area exhibit better crystallinity with a larger average grain size compared to that formed on the regions covered by SiO2 5 outside the pixel area (Extended Data Fig. 3 and Extended Data Fig. 4). Besides, the perovskite formed in devices employing the localised contact scheme shows improved PL uniformity and lifetimes (Extended Data Fig. 5a-c). High-performance NIR micro-PeLEDs with a wide range of pixel sizes are achieved based on the aforementioned fabrication processes (Fig. 2a and Extended Data Fig. 5d). In contrast, emissive pixels with unsatisfactory qualities are obtained without using the localised contact scheme (Extended Data Fig. 5e). Consistent and uniform EL is observed in the NIR micro-PeLEDs (Fig. 2b and Extended Data Fig. 6a-f). Performance data of the devices with various characteristic pixel lengths (Lch=Area1/2) are presented in Fig. 2c-f. The micro-PeLEDs generally exhibit improved voltage tolerance, higher radiance, suppressed EQE roll-off and better spectral stability with reduced pixel sizes (Fig. 2c-f and Supplementary Fig. 1). Remarkably, the micro-PeLEDs show consistently high average EQEs of ~20% across a wide range of characteristic pixel lengths (from ~650 μm to 3.5 μm), without any indication of performance deficit at the microscale (Fig. 2e,f). Further, we extend our investigation to the properties of nano-PeLEDs. SEM images (Fig. 2g and Extended Data Fig. 7a) show arrays of nano-PeLEDs with characteristic lengths of 890, 440 and 90 nm before the deposition of CTL and emissive layer. The smallest nano-PeLEDs, with a characteristic pixel length of ~90 nm, are smaller than the state-of-the-art nano-LEDs based on III-V semiconductors24,25. The reduced pixel sizes enable the demonstration of ultrahigh pixel densities of up to 127,000 PPI, which is record-high amongst all types of LEDs (Supplementary Table 1 and Supplementary Table 2). Perovskite formed in the nanoscale pixel areas exhibits notably improved grain density compared to that in the surrounding insulating areas (Extended Data Fig. 3, Extended Data Fig. 4 and Extended Data Fig. 7), in line with the case of micro-PeLEDs. The microscopic images of the working nano-PeLEDs with a characteristic pixel length of 890, 440 and 90 nm are presented in Extended Data Fig. 7b-d. The diffraction-limited resolution of optical microscopy prevents a clear observation of the nano-PeLED pixel arrays. 6 Fig. 2 | Device characteristics of NIR micro- and nano-PeLEDs. a, Images of working NIR micro-PeLEDs with pixel sizes of 200, 100, 50, 30, 20 and 10 μm. b, EL spectra of NIR micro-PeLEDs. Inset shows the image of a working micro- PeLED array with a characteristic pixel length of 10 μm (driving voltage: 2.5 V). Scale bar: 200 μm. c, Current density- voltage characteristics. d, Radiance-voltage curves. e, EQE-radiance data. f, EQE-current density curves. g, SEM images of pixel areas in the nano-PeLEDs (measured after step (viii) in Fig. 1a). h, Radiance-voltage data of nano-PeLEDs. i, EQE-current density curves of nano-PeLEDs. j, EQE versus characteristic pixel length plot for NIR micro-and nano- PeLEDs. The solid diamonds correspond to the experimentally measured EQE data of the micro- and nano-PeLEDs. The red curves are the Gaussian fitting to the EQE distributions. The box plots and Gaussian fitting were generated from the EQE data, indicating the mean (empty squares), standard deviation, lower quartile (25%), median (50%), upper quartile (75%), interquartile range (25-75%), and maximum/minimum (crosses) of the data. The background colours of cyan, light blue and pink in panel j denote lengthscale regimes corresponding to mini-PeLEDs, micro-PeLEDs and nano-PeLEDs, respectively. 7 The nano-PeLEDs exhibit decent radiance and EQEs under intense charge injection (Figs. 2h & 2i and Supplementary Fig. 2). We note that the peak EQEs of the nano-PeLEDs reduce with decreasing pixel sizes, in contrast to the scenario of our micro-PeLEDs for which the pixel size has little effects on efficiency. The EQE reduction of the nano-PeLEDs may be related to the increased sidewall/pixel area ratio (Extended data Fig. 2f) and the poorer crystallinity of the perovskite formed on pixel areas whose feature sizes are comparable to the grain size32,44 (Extended Data Fig. 7). The effects of downscaling on the peak EQEs of our PeLEDs with characteristic pixel length ranging from 650 μm to ~90 nm are presented in Fig. 2j. It is encouraging to see that the average EQEs are maintained at around 20% across a wide range of characteristic lengths (from 650 to 3.5 μm). Our approach leads to a remarkably small L50 of 180 nm, the pixel length at which the average EQE drops to 50% of that of larger pixels (Fig. 2j). Moreover, lead-free (CsSnI3-based) micro-PeLEDs also exhibit size-insensitive peak EQEs and low EQE roll-off across a wide range of pixel lengths (100 to 3.5 μm) (Supplementary Figs. 3-5), consistent with the downscaling behaviors of lead-based PeLEDs. These results indicate the exceptional advantages of micro/nano-PeLEDs in light-emitting applications. The devices show decent transient response under pulsed voltages, indicating their potential in high-framerate displays and optical communications (Supplementary Figs. 6-7)45,46. Additionally, the downscaling of green, red and sky-blue PeLEDs employing localised contacts is demonstrated. Images of working green and red micro-PeLEDs with characteristic pixel lengths ranging from ~200 μm to ~10 μm are presented in Fig. 3a-b. Arrays of green micro-PeLEDs with various sizes show decent uniformity across different pixels (Extended Data Fig. 8a-d). The spectral stability and uniformity of the EL emission from the green and red micro-PeLEDs are observed through optical microscopy (Fig. 3c, Extended Data Fig. 8e,f,i and Extended Data Fig. 9a). As shown in Fig. 3d-e, the average EQEs of the green micro/nano-PeLEDs are maintained at ~20% for characteristic pixel lengths of 650 μm to 5 μm (Supplementary Figs. 8-10 and Supplementary Table 3), and reduced to around 10% for pixel lengths of 890 nm to ~90 nm (L50 = ~440 nm), showing a scaling effect consistent with that of the NIR micro/nano-PeLEDs. Similar scaling effect was observed for red and sky-blue PeLEDs (Extended Data Fig. 9, Extended Data Fig. 10 and Supplementary Figs. 11-14). The devices show minimum EQE roll-off at high brightness. The peak luminance of green micro-PeLEDs consistently exceeds 300,000 cd m-2 with EQEs of >10% regardless of pixel sizes, showing the potential for AR/VR applications20,47,48. 8 Fig. 3 | Device characteristics of green and red micro/nano-PeLEDs. a-b, Optical microscopic images of working green/red micro-PeLEDs with pixel lengths ranging from 200 μm to ~10 μm. The green (a) and red (b) micro-PeLEDs were driven at 3.0 V and 3.2 V, respectively. Scale bar, 100 μm. c, One-dimensional EL spectral distribution in green micro-PeLEDs. d, EQE-luminance data of green micro-PeLEDs. e, EQE versus characteristic pixel length plot for green micro/nano-PeLEDs. The box plots and Gaussian fitting in e was generated from the EQE data, indicating the mean (empty squares), standard deviation, lower quartile (25%), median (50%), upper quartile (75%), interquartile range (25-75%), and maximum/minimum (crosses) of the data. The background colours of cyan, light blue and pink in panel e denote lengthscale regimes corresponding to mini-PeLEDs, micro-PeLEDs and nano-PeLEDs, respectively. Operational stability tests were conducted for green micro-PeLEDs under current densities of 10 mA cm-2 and 50 mA cm-2. The T50 lifetimes measured are comparable to or exceeding those reported for stable green PeLEDs under similar current densities37,38,49,50. As shown in Fig. 4a and b, the T50 lifetimes increase as the pixel size decreases, reaching 40 h (10 mA cm-2) and 14 h (50 mA cm-2) with a characteristic pixel length of 3.5 μm. Similar trends are found in the T50 lifetime versus characteristic pixel length plots for both current densities (Fig. 4c), indicating the positive effect of downscaling on the operational stability of micro-PeLEDs. To demonstrate the potential of our devices in display applications, a prototypical active-matrix micro-PeLED display based on a commercial thin-film transistor (TFT) array is developed (Methods and Extended Data Fig. 11). The micro-PeLED panel is capable of displaying complex images (Fig. 4d-f) and video clips (Supplementary Videos 1 and 2). 9 Fig. 4 | Device stability measurements and a prototypical active-matrix micro-PeLED display. a-b, Operational stability tests for green micro-PeLEDs under constant current densities of 10 mA cm-2 and 50 mA cm-2 c, T50 lifetimes as functions of characteristic pixel length for green micro-PeLEDs under constant current densities of 10 mA cm-2 and 50 mA cm-2. d-f, Images from a prototypical active-matrix micro-PeLED display powered by a commercial TFT array (pixel dimensions: 70 μm ´ 95 μm). Scale bar: 5 mm. The micro/nano-PeLEDs demonstrated in this work show size-insensitive, high average peak EQEs across a broad range of characteristic pixel lengths (or PPI), in comparison to other classes of LEDs (Fig. 5a). Importantly, our micro/nano-PeLEDs generally show advantages at characteristic pixel lengths of <10 μm, where state-of-the-art III-V LEDs are found to suffer from significant size- dependent efficiency losses. The micro/nano-PeLEDs exhibit the narrowest EL spectra across all classes of miniaturised LEDs, advantageous for next-generation display applications requiring high colour purity (Fig. 5b and Supplementary Table 4). 10 Fig. 5 | EQEs and spectral purity for different classes of miniaturised LEDs. a, EQE versus characteristic pixel length (or PPI) plots for different classes of LEDs. To facilitate an unbiased comparison, the average EQEs from this work are used, while for other studies peak EQE data are used. b, Full-width at half maximum of EL (FWHMEL) for the PeLEDs reported in this work compared to those of other classes of LEDs (with characteristic pixel lengths < 100 μm). Conclusion In summary, we have explored the downscaling of micro- and nano-PeLEDs with characteristic pixel lengths from hundreds of µm to ~90 nm. The lithographically patterned devices employ localised contacts that define the active pixels and prevent non-radiative losses at the pixel boundaries. For our NIR and green devices, average EQEs are maintained at ~20% across a wide range of characteristic lengths (650 to 3.5 μm), exhibiting minimum efficiency reduction upon downscaling. The L50 of these devices are found to be ~180 nm and ~440 nm for the NIR and green devices, respectively. Nano- PeLEDs with characteristic pixel lengths of down to ~90 nm are demonstrated, representing some of the smallest LEDs reported. This enables a record-high pixel density of 127,000 PPI across all types of LED arrays. Besides, a prototypical active-matrix micro-PeLED display based on a commercial TFT array is developed. The micro/nano-PeLEDs generally show advantages at characteristic pixel lengths of <10 μm, where state-of-the-art III-V LEDs are found to suffer from significant size- dependent efficiency losses. Our work highlights the potential of micro- and nano-PeLEDs as a next- generation light source with unprecedented scalability. 11 Methods Materials Poly (9, 9-dioctylfluorene-co-N-(4-(3-methylpropyl)) diphenylamine) (TFB, average molecular weight, ~50,000 g mol-1) was purchased from American Dye Source. Chlorobenzene (extra dry, 99.8%), octane (extra dry, >99%), ethanol (extra dry, 99.5%), N, N-dimethylformamide (DMF, 99.5%), Dimethyl sulfoxide (DMSO, HPLC grade) and ethyl acetate (HPLC grade) were purchased from J&K Chemical Ltd. Formamidinium iodide (FAI, 99.99%), lead iodide (PbI2, 99.999%), 2,4,6-tris[3- (diphenylphosphinyl) phenyl]-1,3,5-triazine (PO-T2T, 99.99%), MoOx (99.9%), were purchased from Xi’an Polymer Light Technology Corp. Formamidine hydrobromide (FABr, 99.99%) was purchased from Tokyo Chemical Industry Co., Ltd. Guanidinium bromide (GABr, 98%), 2-methacryloyloxyethyl phosphorylcholine (MPC, contains ≤100 ppm MEHQ as inhibitor, 97%), Cesium bromide (CsBr, 99.99%), lead bromide (PbBr2, 99.999%), LiF (99.99%), Sulfonamide (SFA), β-alaninamide hydrochloride (BAH), cesium iodide (CsI, 99.999%), lead chloride (PbCl2, 99.999%), lead chloride (PbCl2, 99.999%), rubidium bromide (RbBr, 99.99%), 4,7,10-trioxa-1,13-tridecanediamin (TTDDA), poly(9-vinylcarbazole) (PVK, average Mn 25,000–50,000) and 1,3,5-tris(1-phenyl-1H-benzimidazol- 2-yl)benzene (TPBi) were purchased from Sigma-Aldrich. Formamidinium bromide (FABr) and NiOx were purchased from Xi’an Yuri Solar Co., Ltd. All materials were used as received without further purification. Fabrication of near-infrared PeLEDs without localised contacts A 135-nm-thick ITO electrode was sputtered via RF magnetron to achieve a sheet resistance of 30 Ω per square. The ITO was etched using a laser etching system with a 1064 nm laser source to produce the stripe patterns. The ITO substrates were sequentially cleaned using deionized water, acetone, and isopropanol under ultrasonication. Following the cleaning steps, the substrates were exposed to UV- ozone for 15 min. Colloidal ZnO nanoparticles were spin-coated onto the patterned substrates at 5000 rpm for 45 s and annealed in air at 150 °C for 10 min. Next, PEIE solution (0.03 wt% in isopropanol) was spin-coated onto the ZnO surface at 5000 rpm for 45 s followed by annealing at 100 °C for 10 min. Subsequently, the substrates were transferred into a N2-filled glovebox. The perovskite precursor solution was prepared by dissolving 3.4 mg CsI, 42.5 mg FAI, 60 mg PbI2, and 6 mg β-alaninamide hydrochloride (BAH) in 1 mL of DMF. The perovskite solution was filtered with 0.22-μm filters before spin-coating. The perovskite films were prepared by spin-coating the precursor solution onto the PEIE- treated ZnO films at 5000 rpm for 70 s, followed by annealing at 96 °C for 10 min. TFB (12 mg mL-1 in chlorobenzene) was spin-coated at 4000 rpm for 45 s. Then the MoOx were deposited using a thermal evaporation system through a shadow mask under a base pressure of 4×10-4 Pa. Subsequently, the 12 width of the Au electrode was defined by another metal mask. The effective area of the PeLEDs was determined by the vertical intersection of the bottom ITO and the top Au electrode. The widths of bottom ITO and top Au range from 800 μm to 50 μm. Due to the precision limitations of the ITO etching system and the metal mask preparation equipment, a noticeable deviation occurs at the edges of the fabricated PeLEDs. All devices were encapsulated with UV epoxy (NOA81, Thorlabs)/cover glass before subsequent measurements. Preparation of patterned substrates for micro- and nano-PeLEDs with localised contacts A 135-nm-thick ITO electrode was sputtered via RF magnetron to achieve a sheet resistance of 30 Ω per square. The ITO was etched using a laser etching system with a 1064 nm laser source to produce the stripe patterns. Patterned ITO glass substrates were sequentially cleaned using deionized water, acetone, and isopropanol under ultrasonication. A 130-nm-thick SiO2 layer was deposited onto the ITO substrates by plasma enhanced chemical vapor deposition (PECVD, HQ-8B). To prepare the patterned substrates for the micro-PeLEDs, a positive photoresist layer was spin- coated onto the SiO2 layer and prebaked at 100 °C for 10 min. The photoresist was exposed under a 365-nm UV lamp for 4–6 s, followed by a development step in photoresist developer for 30 s. During the photolithography process, precise alignment was achieved using a 10×100 μm2 marker at the edge of the photomask and the patterned substrates. The patternes were formed by etching off the 130-nm- thick unprotected SiO2 using reactive-ion etching (RIE). In the processing of RIE, an excessive 5 nm of ITO was etched off to ensure that all SiO2 was completely removed from the patterned structures. Then the patterned structures were immersed in N-methyl pyrrolidone (NMP) solution until the residual photoresist was removed. To prepare the patterned substrates for the nano-PeLEDs, nanohole arrays with the designed diameters and a set duty cycle of 50% were fabricated using a focused-ion-beam (FIB) system (Auriga 40, Carl Zeiss) under a high vacuum of 9.05×10-7 mbar. In the patterned ITO covered by SiO2, the central region of ITO was chosen as the etching area. For nano arrays with pixel diameters of 1000 nm and 500 nm (corresponding to characteristic pixel lengths of 890 nm and 440 nm), the FIB system was operated at a beam current of 10 pA at 30 kV. For nano arrays with diameters of 200 nm and 100 nm (corresponding to characteristic pixel lengths of 180 nm and 90 nm), the FIB system was operated at a beam current of 5 pA at 30 kV. The optimized thickness was achieved by etching off the unprotected 130-nm-thick SiO2, exposing the underlying ITO. A 5-nm SiO2 passivation layer for the micro/nano- patterned substrates was deposited by magnetron sputtering (Discovery-635, DENTON). During the fabrication process, the thicknesses of etched SiO2 on ITO substrates were determined by a Bruker DEKTAK-XT profilometer. 13 Fabrication of near-infrared micro- and nano-PeLEDs with localised contacts The perovskite precursor solution was prepared by dissolving 3.4 mg CsI, 42.5 mg FAI, 60 mg PbI2, and 6 mg β-alaninamide hydrochloride (BAH) in 1 mL of DMF. The perovskite solution was filtered with 0.22-μm filters before spin-coating. The micro/nano-patterned substrates prepared using the localised contact method were sequentially cleaned using deionized water, acetone, and isopropanol under ultrasonication. Following the cleaning steps, the substrates were exposed to UV-ozone for 15 min. Colloidal ZnO nanoparticles were spin-coated onto the micro/nano-patterned substrates at 5000 rpm for 45 s and annealed in air at 150 °C for 10 min. Next, PEIE solution (0.03 wt% in isopropanol) was spin-coated onto the ZnO surface at 5000 rpm for 45 s followed by annealing at 100 °C for 10 min. Subsequently, the substrates were transferred into a N2 glovebox. The perovskite films were prepared by spin-coating the precursor solution onto the PEIE-treated ZnO films at 5000 rpm for 70 s, followed by annealing at 96 °C for 10 min. TFB in chlorobenzene (12 mg mL-1) was spin-coated at 4000 rpm for 45 s. MoOx was deposited using a thermal evaporation system through a shadow mask under a base pressure of 4×10-4 Pa. Subsequently, the Au (100 nm) electrode fully covering the localised contact region was evaporated through another shadow mask. The active area of the micro/nano-PeLEDs was defined by the localised contact region. These localised contact regions are sufficiently far from the boundaries of the bottom ITO electrode and the top Au electrode. All devices were encapsulated with UV epoxy (NOA81, Thorlabs)/cover glass before subsequent measurements. Fabrication of red micro/nano-PeLEDs The mixed halide FA0.5Cs0.5PbIxBr3-x precursor solution was prepared by dissolving CsI, PbI2, FABr, FAI, and sulfonamide (SFA) in DMF with a molar ratio of 0.5:1:(0.5−x):x:y (x = 0-0.25, y = 0-0.06). The concentration of Pb2+ in the precursor solution was 0.11 mol L-1. All the precursor solutions were stirred at room temperature in an N2-filled glovebox for about 2 h. The solution was filtered with PTFE filters (0.22 μm) before use. The micro/nano-patterned substrates were sequentially cleaned using deionized water, acetone, and isopropanol under ultrasonication. Following the cleaning steps, the substrates were exposed to UV-ozone for 15 min. Synthesized ZnO nanoparticles were spin-coated onto the substrates at 5000 rpm for 45 s, followed by annealing at 150 °C for 10 min. Subsequently, PEIE (0.04 wt% in isopropanol) was spin-coated onto the ZnO layer at 5000 rpm for 45 s, followed by annealing at 100 °C for 10 min. After cooling, the substrates were transferred to a N2-filled glovebox. To fabricate the perovskite films, the precursors were spin-coated onto the PEIE-modified ZnO substrates at 5000 rpm for 60 s, followed by annealing at 100 °C for 10 min. Next, a TFB solution (12 mg mL-1 in chlorobenzene) was spin-coated at 4000 rpm for 45 s. Finally, MoOx (15 nm) were 14 evaporated at 0.2 Å s-1 through a shadow mask under high vacuum. Subsequently, the Au (100 nm) electrode fully covering the localised contact region was evaporated through another shadow mask. Fabrication of green micro/nano-PeLEDs The perovskite precursor solution was prepared by dissolving 21.3 mg CsBr, 17.5 mg FABr, 73.4 mg PbBr2, 2.8 mg GABr, and 5 mg MPC into 1 mL DMSO. The solution was stirred at 50 °C for 6 h before use. The perovskite solution was filtered with 0.22-μm filters before spin-coating. The molar concentration of PbBr2 was 0.2 M. Following the cleaning steps, the substrates were treated with UV- Ozone for 20 min. Nickle oxide nanoparticle dispersion (20 mg mL-1 in deionized water, ultrasonically treated for 5 min) was spin-coated onto the patterned substrates at 4000 rpm for 60 s and annealed at 150 ℃ for 30 min in the fume hood. Then the substrates were transferred into a N2-filled glove box, and TFB solution (10 mg mL-1 in chlorobenzene) was spin-coated onto the substrates at 4000 rpm for 60 s, followed by annealing at 150 ℃ for 30 min. After cooling to room temperature, SrF2 (1 nm) was thermally evaporated onto the TFB layer. The perovskite precursor solution was spin-coated at 4500 rpm for 60 s, followed by annealing at 100 ℃ for 10 min. PO-T2T (~40 nm) was evaporated through a shadow mask at a pressure < 2×10-4 Pa. Subsequently, the LiF (1 nm) and Al (100 nm) electrodes fully covering the localised contact regions were evaporated through another shadow mask. Fabrication of sky-blue micro-PeLEDs The perovskite precursor (concentration of Pb2+: 0.2 M) was prepared by dissolving 51 mg CsBr, 44 mg PbBr2, 22 mg PbCl2, 5 mg FABr, 3 mg RbBr and 4 mg TTDDA in 1 mL of DMSO and stirred for 6 h. The micro-patterned substrates prepared using the localised contact method were sequentially cleaned by deionized water, acetone, and isopropanol (IPA) in an ultrasonic bath for 15 min. The clean substrates were then treated by UV-ozone. NiOx nanodispersion (10 mg mL-1 in deionized water) was spin-coated in air at 3000 rpm for 30 s, followed by baking at 150 °C for 20 min in air. The substrates were then transferred into a N2-filled glovebox. PVK (10 mg mL−1 in chlorobenzene) was spin-coated at 3000 rpm, followed by thermal annealing at 150 °C for 30 min. Then the perovskite precursor was spin-coated at 3000 rpm for 60 s. Subsequently, the perovskite film was treated by vapour-assisted crystallization and annealed at 80 °C for 10 min. Finally, TPBi (55 nm), LiF (1.5 nm) and Al (100 nm) were sequentially evaporated at a pressure < 2×10-4 Pa. 15 Fabrication of the active-matrix micro-PeLED display The TFT array substrate (LinkZill, LC-T-G256-GDQ2) was sequentially cleaned by deionized water, acetone, and isopropanol (IPA) in an ultrasonic bath for 15 min. The perovskite precursor solution was prepared by dissolving 21.3 mg CsBr, 17.5 mg FABr, 73.4 mg PbBr2, 2.8 mg GABr, and 5 mg MPC into 1 mL DMSO. The solution was stirred at 50 °C for 6 h before use. The perovskite solution was filtered with 0.22-μm filters before spin-coating. The molar concentration of PbBr2 was 0.2 M. Following the cleaning steps, the substrate was treated with UV-Ozone for 20 min. Nickle oxide nanoparticle dispersion (20 mg mL-1 in deionized water, ultrasonically treated for 5 min) was spin- coated onto the substrate at 4000 rpm for 60 s and annealed at 150 ℃ for 30 min in the fume hood. Then the substrate was transferred into a N2-filled glove box, and TFB solution (10 mg mL-1 in chlorobenzene) was spin-coated onto the substrate at 4000 rpm for 60 s, followed by annealing at 150 ℃ for 30 min. After cooling to room temperature, SrF2 (1 nm) was thermally evaporated onto the TFB layer. The perovskite precursor solution was spin-coated at 4500 rpm for 60 s, followed by annealing at 100 ℃ for 10 min. PO-T2T (~40 nm) was evaporated through a shadow mask (designed for the TFT array) at a pressure < 2×10-4 Pa. Subsequently, the LiF (1 nm) and Al (100 nm) electrodes were evaporated through a shadow mask designed for the TFT array. Characterization of micro- and nano-PeLEDs The EL and EQE data of the micro- and nano-PeLEDs were measured using an integrating sphere fiber-coupled to an Ocean Optics QEPro spectrometer in a N2-filled glovebox. The intergrating sphere was calibrated using a standard Vis–NIR light source (HL-3P-INT-CAL plus, Ocean Optics). The J– V curves of PeLEDs were obtained using a Keithley 2400 sourcemeter unit, with a voltage stepping rate of 0.1 V s−1. The current density (J) of a micro/nano-PeLED was calculated by: the total current / (the nominal area of each active pixel × the total number of pixels in an array). The nominal areas of the pixels were found to be in close agreement with microscopic observations. Operational stability measurements for green micro-PeLEDs The operational stability tests were performed in a N2-filled glovebox at ambient temperature (25 ± 5 °C) using a multichannel LED lifetime-testing system (Crysco). SEM and STEM measurements Scanning electron microscopy (SEM) measurements were carried out using a field-emission scanning electron microscope (HITACHI SU8010). The surface morphology and roughness of the samples were collected by atomic force microscope (AFM) (Bruker Dimension FastScan) with contact mode. The 16 near-infrared-emitting perovskite films used in the measurements were deposited on glass/ITO/patterned substrate/PEIE modified ZnO. Scanning transmission electron microscopy (STEM) and high-angle annular dark-field scanning transmission electron microscopy (HAADF- STEM) results were obtained with an aberration-corrected FEI Titan G2 80–200 ChemiSTEM instrument. TCSPC measurements The PL intensity and lifetime were obtained using a time-correlated single-photon counting (TCSPC) setup (PicoQuant, MicroTime 200). A 400-nm 500-kHz pulsed laser was focused onto the sample with a ´50 objective. The excitation energy density is ~5 nJ cm-2. FLIM measurements Fluorescence lifetime imaging microscopy (FLIM) measurements were conducted with Leica Stellaris 8 FALCON. The wavelength of the laser for the excitation of samples was 450 nm. The detection range was 780-830 nm. The scan speed was 10 Hz. Temporal response measurements Short square voltage pulses were generated by a DG1062Z function generator and applied to the PeLEDs. A 50-Ω resistor was placed in series with the devices. Transient voltages were measured using a four-channel oscilloscope (Tektronix Oscilloscope MDO34) with a bandwidth of 1 GHz and samling rate of 5 GHz/s. The transient current was calculated from the voltage across the resistor in series with the PeLEDs. The bandwidth of the PeLEDs were driven by a square voltage and the average EL intensities were measured using an optical power meter (Newport, 818-UV/DB). To reliably measure the temporal EL response of the devices, the lengths of the cable and wiring employed in the experiments were kept at minimum; the devices and oscilloscope were connected using a fast Bayonet Neil-Concelman (BNC) connector to minimize the influence of the inductance of the circuit. The capacitance-frequency (C-F) data were obtained using a Keithley 4200A semiconductor parametric analyser. Imaging of working PeLEDs An optical microscope was used to observe the working PeLEDs. The EL spectra of the devices were measured using an Ocean Optics QEPro spectrometer. A hyperspectral microcoscope was used to acquire additional images of the devices. 17 References 1. Cao, W. et al. The future transistors. Nature 620, 501-515 (2023). 2. Lundstrom, M. S. & Alam, M. A. Moore's law: The journey ahead. Science 378, 722-723 (2022). 3. Kim, T. I. et al. Injectable, cellular-scale optoelectronics with applications for wireless optogenetics. Science 340, 211-216 (2013). 4. Park, S. I. et al. Soft, stretchable, fully implantable miniaturized optoelectronic systems for wireless optogenetics. Nat. Biotechnol. 33, 1280-1286 (2015). 5. Behrman, K. & Kymissis, I. Micro light-emitting diodes. Nat. Electron. 5, 564-573 (2022). 6. Jin, S. et al. GaN microdisk light emitting diodes. Appl. Phys. Lett. 76, 631-633 (2000). 7. Jiang, H. et al. III-nitride blue microdisplays. Appl. Phys. Lett. 78, 1303-1305 (2001). 8. Huang, Y. et al. Mini-LED, Micro-LED and OLED displays: present status and future perspectives. Light Sci. Appl. 9, 105 (2020). 9. Shin, J. et al. Vertical full-colour micro-LEDs via 2D materials-based layer transfer. Nature 614, 81-87 (2023). 10. Amano, H. et al. P-type conduction in Mg-doped gan treated with low-energy electron-beam irradiation (Leebi). Jpn. J. Appl. Phys. 28, L2112-L2114 (1989). 11. Nakamura, S. et al. Candela-class high-brightness InGaN/AlGaN double-Heterostructure blue- light-emitting diodes. Appl. Phys. Lett. 64, 1687-1689 (1994). 12. Ponce, F. A. & Bour, D. P. Nitride-based semiconductors for blue and green light-emitting devices. Nature 386, 351-359 (1997). 13. Meng, W. et al. Three-dimensional monolithic micro-LED display driven by atomically thin transistor matrix. Nat. Nanotechnol. 16, 1231-1236 (2021). 14. Baek, W. et al. Ultra-low-current driven InGaN blue micro light-emitting diodes for electrically efficient and self-heating relaxed microdisplay. Nat. Commun. 14, 1386 (2023). 15. Meng, T. et al. Ultrahigh-resolution quantum-dot light-emitting diodes. Nat. Photon. 16, 297-303 (2022). 16. Zhao, J. et al. Large-area patterning of full-color quantum dot arrays beyond 1000 pixels per inch by selective electrophoretic deposition. Nat. Commun. 12, 4603 (2021). 17. Fan, J. & Qian, L. Quantum dot patterning by direct photolithography. Nat. Nanotechnol. 17, 906- 907 (2022). 18. Liu, Z. et al. Micro-light-emitting diodes with quantum dots in display technology. Light Sci. Appl. 9, 83 (2020). 19. Deng, T. et al. Pixelated phosphor converter for laser-driven adaptive lighting. Laser Photon. Rev. 17, 2300240 (2023). 18 20. Park, J. et al. Electrically driven mid-submicrometre pixelation of InGaN micro-light-emitting diode displays for augmented-reality glasses. Nat. Photon. 15, 449-455 (2021). 21. Lu, B. et al. Dennard scaling in optoelectronics: scientific challenges and countermeasures of micro-LEDs for displays. Laser Photon. Rev. 16, 2100433 (2022). 22. Zhuang, Z. et al. Investigation of InGaN-based red/green micro-light-emitting diodes. Opt. Lett. 46, 1912-1915 (2021). 23. Dussaigne, A. et al. Full InGaN red (625 nm) micro-LED (10 μm) demonstration on a relaxed pseudo-substrate. Appl. Phys. Express 14, 092011 (2021). 24. Sheen, M. et al. Highly efficient blue InGaN nanoscale light-emitting diodes. Nature 608, 56-61 (2022). 25. Liu, X. et al. N-polar InGaN nanowires: breaking the efficiency bottleneck of nano and micro LEDs. Photonics Res. 10, 587-593 (2022). 26. Tan, Z.-K. et al. Bright light-emitting diodes based on organometal halide perovskite. Nat. Nanotechnol. 9, 687-692 (2014). 27. Cho, H. et al. Overcoming the electroluminescence efficiency limitations of perovskite light- emitting diodes. Science 350, 1222-1225 (2015). 28. Wang, N. et al. Perovskite light-emitting diodes based on solution-processed self-organized multiple quantum wells. Nat. Photon. 10, 699-704 (2016). 29. Yuan, M. et al. Perovskite energy funnels for efficient light-emitting diodes. Nat. Nanotechnol. 11, 872-877 (2016). 30. Liu, X. et al. Metal halide perovskites for light-emitting diodes. Nat. Mater. 20, 10-21 (2021). 31. Zhao, B. et al. High-efficiency perovskite-polymer bulk heterostructure light-emitting diodes. Nat. Photon. 12, 783-789 (2018). 32. Cao, Y. et al. Perovskite light-emitting diodes based on spontaneously formed submicrometre- scale structures. Nature 562, 249-253 (2018). 33. Lin, K. et al. Perovskite light-emitting diodes with external quantum efficiency exceeding 20 percent. Nature 562, 245-248 (2018). 34. Chiba, T. et al. Anion-exchange red perovskite quantum dots with ammonium iodine salts for highly efficient light-emitting devices. Nat. Photon. 12, 681-687 (2018). 35. Sun, Y. et al. Bright and stable perovskite light-emitting diodes in the near-infrared range. Nature 615, 830-835 (2023). 36. Guo, B. et al. Ultrastable near-infrared perovskite light-emitting diodes. Nat. Photon. 16, 637-643 (2022). 19 37. Kim, J. et al. Ultra-bright, efficient and stable perovskite light-emitting diodes. Nature 611, 688- 694 (2022). 38. Chen, W. et al. Highly bright and stable single-crystal perovskite light-emitting diodes. Nat. Photon. 17, 401-407 (2023). 39. Hsu, C. et al. Efficient mini/micro-perovskite light-emitting diodes. Cell Rep. Phys. Sci. 2, 100582 (2021). 40. Yang, X. et al. Towards micro-PeLED displays. Nat. Rev. Mater. 8, 341–353 (2023). 41. Liu, D. et al. Direct optical patterning of perovskite nanocrystals with ligand cross-linkers. Sci. Adv. 8, eabm8433 (2022). 42. Liu, D. et al. Nondestructive direct optical patterning of perovskite nanocrystals with carbene- based ligand cross-linkers. ACS Nano 18, 6896-6907 (2024). 43. Kim, M. et al. High-performance blue perovskite films and micro-arrays for light-emitting diodes with ionic liquid interlayer. Adv. Funct. Mater. 34, 2309436 (2023). 44. Jastrzebska-Perfect, P. et al. On-site growth of perovskite nanocrystal arrays for integrated nanodevices. Nat. Commun. 14, 3883 (2023). 45. Ren, A. et al. High-bandwidth perovskite photonic sources on silicon. Nat. Photon. 17, 798-805 (2023). 46. Bao, C. et al. Bidirectional optical signal transmission between two identical devices using perovskite diodes. Nat. Electron. 3, 156-164 (2020). 47. Xiong, J. et al. Augmented reality and virtual reality displays: emerging technologies and future perspectives. Light Sci. Appl. 10, 216 (2021). 48. Kim, S. et al. Primary color generation from white organic light-emitting diodes using a cavity control layer for AR/VR applications. Org. Electron. 87, 105938 (2020). 49. Ding, S. et al. Phase dimensions resolving of efficient and stable perovskite light-emitting diodes at high brightness. Nat. Photon. 18, 363–370 (2024). 50. Ma, D. et al. Distribution control enables efficient reduced-dimensional perovskite LEDs. Nature 599, 594-598 (2021). Acknowledgements This work was supported by the National Key Research and Development Program of China (2022YFA1204800), National Natural Science Foundation of China (NSFC) (62274144, 62005243), the Zhejiang Provincial Government, Natural Science Foundation of Zhejiang Province (LR21F050003), Fundamental Research Funds for the Central Universities, and Zhejiang University 20 Education Foundation Global Partnership Fund. S.D.S. acknowledges support from the Royal Society and Tata Group (UF150033). K.J. acknowledges the Royal Society studentship. The authors acknowledge funding support from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (HYPERION, grant agreement number 756962). We are grateful to Dr. Linrun Feng and LinkZill Technology (Hangzhou) for the provision of the TFT array which powered the active-matrix micro-PeLED display. We thank Wei Wang (Zhejiang Univ.) for his assistance with FIB and SEM measurements, and Dr. Peizhen Xu and Dr. Liu Yang (Zhejiang Univ.) for helpful discussions. We thank Dr. Linjie Dai (Cambridge Univ.) for his assistance with the hyperspectral microscopic experiments. We acknowledge the Center of Electron Microscopy, Zhejiang University, for their assistance with the TEM, FIB and EDX experiments. Dr. Jiabao Sun of ZJU Micro-Nano Fabrication Center is acknowledged for his professional and enthusiastic technical assistance. The authors thank Dr. Lingyu Xiao from the Instrumentation and Service Center for Molecular Sciences at Westlake University for the assistance with the FLIM measurements. The authors thank the Core Facilities, State Key Laboratory of Extreme Photonics and Instrumentation, Zhejiang University for technical support. We thank Minhui Yu, Yuzhen Zhao and Xiyan Yang for their administrative support. Author contributions Y.L. planned the research and designed the experiments under the guidance of D.D. and B.Z.. Y.L. fabricated the micro- and nano-PeLEDs with Y.W., Y.Y. and Z.R., and carried out the device characterizations. Y.L. and Y.W. fabricated the near-infrared micro- and nano-PeLEDs. Y.L. and Y.W. carried out the optical microscopy, SEM, and transient EL measurements. Y.Y. and S.X. fabricated and characterized the green micro-PeLEDs. Y.Y. carried out the operational stability measurements of green micro-PeLEDs. Y.L. and Y.Y. prepared the green active-matrix micro-PeLED display. Z.R. and Y.W. fabricated and measured the red micro/nano-PeLEDs. W.T. fabricated the CsSnI3-based micro- PeLEDs. Z.L. contributed to the fabrication of the blue micro-PeLEDs. K.J. performed the hyperspectral microscopy experiments under the supervision of S.D.S. B.Y. contributed to device analyses. Y.Y. carried out the TCSPC measurements. Y.G., S.Z., K.Z. and G.Z. contributed to experiments and analyses. Y.L., Y.W., Y.Y. and Z.R. wrote the initial draft of the manuscript, which was revised by D.D. and B.Z. All authors contributed to the work and commented on the paper. Competing interests D.D., Y.L., B.Z. and Y.W. are inventors on CN patent application: 202410198153.5. The remaining authors declare no competing interests. 21 Data availability The data supporting the findings of this study are available within the paper and its supplementary information. Corresponding authors Dawei Di (daweidi@zju.edu.cn) and Baodan Zhao (baodanzhao@zju.edu.cn). 22 Extended Data Fig. 1 | Substrates fabricated with and without using the localised contact method. a, b, Schematic diagram of devices fabricated without and with the localised contact method, respectively. For devices prepared using the localised contact method, an etched insulating layer is used to define the active area. The sky-blue arrow represents the direction of the current flow. c, d, Top-view SEM images of the pixel boundaries prepared without and with the localised contact method, respectively. Scale bar, 20 μm. Scale bar, 500 nm of d (ii). e, The depth profiles of localised contact area by the profilometer. f, AFM images of the active areas prepared by the localised contact method. 23 Extended Data Fig. 2 | Perovskites on the substrates fabricated with and without using the localised contact method. a, AFM images of the active area with a characteristic length of 30 μm prepared by the localised contact method. b, Roughness of the localised contact area (region (i) in panel a). Scale bar: 4 μm. c, Roughness of the insulated area (region (ii) in panel a). Scale bar, 4 μm. d, SEM images of perovskite samples on substrate fabricated without using the localised contact method. Scale bar, 30 μm. e, SEM images of perovskite samples on substrate prepared by the localised contact method. Scale bar, 30 μm. f, Sidewall/pixel area ratio (%) as a function of characteristic length. 24 Extended Data Fig. 3 | Morphological characteristics of perovskite samples on substrates prepared by the localised contact method. a, Top-view SEM images of the active area of perovskite samples on substrates prepared by the localised contact method. The upper right panel shows the SEM images of the perovskite deposited in the active pixel area. The bottom right panel shows the SEM images of the perovskite deposited across the active pixel region and the insulating region. Scale bar represents 30 μm and 1 μm, respectively. b-c, Distributions of perovskite grain sizes in the active pixel region and the insulating region. 25 Extended Data Fig. 4 | SEM image and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images of the perovskite samples on the substrates prepared by the localised contact method. a, Top-view SEM images of the active area of the perovskite samples on substrates prepared by the localised contact method. Scale bar, 30 μm. HAADF-STEM images of the perovskite material deposited in the insulating area (b) and in the central region of the localised contact area (c). Scale bar, 5 nm. 26 Extended Data Fig. 5 | PL and EL imaging of NIR perovskites on substrates fabricated with and without using the localised contact method. a-b, FLIM images of near-infrared perovskite samples on substrates fabricated without and with the localised contact method, respectively. c, Time-resolved PL decay kinetics of the perovskite in different regions denoted in panels a and b. d-e, Fluorescence optical microscopic images of NIR micro-PeLEDs fabricated with (d) and without (e) the localised contact method with pixel lengths ranging from 200 μm to ~10 μm. 27 Extended Data Fig. 6 | Additional fluorescence measurements of NIR micro-PeLEDs fabricated with and without the localised contact method. a, Normalized EL spectra of micro-PeLEDs using the localised contact method with pixel lengths ranging from 650 μm to ~10 μm at a radiance of 100 W sr-1 m-2. b, Time-resolved PL decay kinetics of perovskite samples on substrates using the localised contact method with various characteristic pixel lengths. Fluence: 5 nJ cm-2. c, EL mapping of the area with 100×100 μm2 of a near-infrared micro-PeLED using the localised contact method. The measured region of 100×100 μm2 is in the center of the micro-PeLED with a characteristic length of 200 μm. The scanning step interval is 1 μm. d, EL spectral information in the selected region in panel c along X axis. e-f, Hyperspectral microscopy images of the micro-PeLEDs fabricated without (e) and with (f) the localised contact method. 28 Extended Data Fig. 7 | Images of NIR and green nano-PeLED arrays fabricated using the localised contact method. a, SEM images of the patterned pixel areas in nano-PeLEDs with characteristic pixel lengths of 890 nm, 440 nm and 90 nm. b, Optical microscopic images of the patterned pixel areas in nano-PeLEDs with characteristic pixel lengths of 890 nm, 440 nm and 90 nm. c, Optical microscopic images of working NIR nano-PeLED arrays with characteristic pixel lengths of 890 nm, 440 nm and 90 nm. d, Optical microscopic images of working green nano-PeLED arrays with characteristic pixel lengths of 890 nm, 440 nm and 90 nm. 29 Extended Data Fig. 8 | Additional measurements for PeLEDs fabricated using the localised contact method. a-d, Optical microscope images of green micro-PeLED arrays with different pixel resolutions of 64 PPI (a), 635 PPI (b), 1270 PPI (c) and 2540 PPI (d). Scale bar, 200 μm. e, EL mapping of the area with 100x100 μm2 of a green micro-PeLED using the localised contact method. The measured region of 100×100 μm2 is in the center of the micro-PeLED with a characteristic length of 200 μm. The scanning step interval is 1 μm. f, EL spectral information of the PeLEDs in the boxed region of panel e. g-h, Maximum luminance versus characteristic pixel length for the green (g) and red (h) PeLEDs. i, Spectral stability of red micro-PeLEDs fabricated using localised contact method. EL spectral is obtained from the centre of a red micro-PeLED with a characteristic length of 200 μm. 30 Extended Data Fig. 9 | Spectral and efficiency data of red micro-/nano-PeLEDs. a, One-dimensional EL spectral distribution in red micro-PeLEDs. b, EQE-luminance data. c, EQE versus characteristic pixel length plot for red micro/nano-PeLEDs. The box plots and Gaussian fitting was generated from the EQE data, indicating the mean (empty squares), standard deviation, lower quartile (25%), median (50%), upper quartile (75%), interquartile range (25-75%), and maximum/minimum (crosses) of the data. The background colours of cyan, light blue and pink denote lengthscale regimes corresponding to mini-PeLEDs, micro-PeLEDs and nano-PeLEDs, respectively. 31 Extended Data Fig. 10 | Efficiency and spectral data of sky-blue micro-PeLEDs. a, EQE versus characteristic pixel length plot for sky-blue micro-PeLEDs. b, EL spectrum of a sky-blue micro-PeLED (peak wavelength: 488 nm). Inset shows the image of a working micro-PeLED array with a characteristic pixel length of 10 μm (driving voltage: 5.5 V). Scale bar: 200 μm. 32 Extended Data Fig. 11 | Structure of the active-matrix micro-PeLED display. The fabrication processes of the active- matrix micro-PeLED display is available in the Methods section. The TFT array was provided by LinkZill Technology (Hangzhou). IGZO denotes indium gallium zinc oxide.