Introduction

Perovskite quantum dots (QDs) have been regarded as star materials for new generation high-definition (HD) displays because of their narrow emission spectra (full width at half maximum (FWHM) < 35 nm), high defect tolerance, and facile solution processability1,2,3,4. These properties enable perovskite QDs to meet the stringent color standards of Rec. 2020, which is critical for ultra-high-definition displays and advanced applications such as augmented/virtual reality (AR/VR). Among the three primary colors, blue emission is the most challenging: it directly dictates overall display color gamut and is particularly difficult to realize with high efficiency, high color purity, and excellent operational stability.

Recent breakthroughs in device engineering have propelled the maximum external quantum efficiency (EQE) of blue perovskite light-emitting diodes beyond 20%, rivaling cadmium-based quantum dot light-emitting diodes (QLEDs)5,6,7,8,9,10,11,12. However, this performance is typically accompanied by poor color purity—most efficient devices exhibit a Commission International de l’Eclairage coordinates of y (CIEy) > 0.1, far below the Rec. 2020 standard for blue. Moreover, devices tuned toward high color purity blue (CIEy < 0.1) often suffer from severe efficiency roll-off under practical luminance (>2000 cd m−²) and short operational lifetimes (T₅₀ < 100 min at 100 cd m−²) due to surface defects, Auger recombination, and charge-injection imbalance13,14,15,16,17,18,19. Therefore, achieving simultaneously high color purity, low roll-off, and long lifetime remains an unresolved bottleneck toward commercial viability.

Efforts to tackle these issues have focused on composition engineering, such as alloying with chloride to push emission toward high color purity blue5,18,19,20,21,22,23, and A-site engineering to passivate surface traps16,17. While these approaches have yielded incremental gains, they often introduce new problems: compositional inhomogeneity leads to spectral instability, while insulating ligands improve passivation but hinder carrier injection and film conductivity. To date, no strategy has simultaneously achieved high color purity (CIEy < 0.1), >20% EQE at high luminance (>5000 cd m−2), and extended operational stability (>100 min at 100 cd m−²) in blue perovskite QLEDs.

Here, we report a multifunctional molecule passivation strategy using 1-ethyl-3-methylimidazolium hexafluorophosphate (EMIMPF₆) to address these long-standing challenges. The [PF₆]− ions bind strongly to cesium sites and lead dangling bonds, thereby decreasing defect-assisted non-radiative recombination and weakening inter-dot coupling. Simultaneously, the [EMIM]⁺ ions interact with bromine dangling bonds, tune the band alignment of perovskite QDs, thereby reducing hole-injection barriers and enhancing charge balance of LEDs under electrical bias. Moreover, their high permittivity suppresses the Auger recombination of QDs, and finally enables the unit of high efficiency, high luminance, and long stability. Compared to the previous work, we improve the operational stability and EQE of high color purity blue perovskite QLEDs; meanwhile, we reduce the efficiency roll-off of blue perovskite QLEDs, achieving an EQE of 20.02% at a brightness of 6441 cd m−2, and when the luminance increases to 9587 cd m−2, it can still maintain an EQE of 18.47%. This work demonstrates that multifunctional molecule passivation can simultaneously optimize optical, electronic, and stability properties of perovskite QDs, providing a generalizable pathway toward ultra-high-definition blue QLEDs that meet commercial display requirements.

Results

We synthesized pure-bromide QDs with strong quantum confinement by reducing their size, thereby ensuring that the emission color coordinates fall within the Rec. 2020 blue region. To passivate dangling bonds caused by dynamic bonding of long-chain ligands and suppress Auger recombination caused by a decrease in QD size, we introduced a multifunctional molecule—EMIMPF₆, using a heterogeneous treatment strategy. In nonpolar solvents, the EMIMPF₆ dissociates into 1-ethyl-3-methylimidazolium ([EMIM]⁺) cations and hexafluorophosphate ([PF₆]−) anions, which selectively interact with bromine and lead/cesium dangling bonds, respectively (Fig. 1a).

Fig. 1: The introduction of the multifunction molecule.
Fig. 1: The introduction of the multifunction molecule.
Full size image

a The schematic diagram of introducing EMIMPF₆ onto the surface of QDs. b, c The adsorption sites of the b [PF₆]− and c [EMIM]⁺ ions on the QD surface. d The adsorption energies of the [PF₆]− and [EMIM]⁺ ions on the QD surface. e–g The e FTIR spectra, f XRD patterns, and g fine XPS of Pb 4f of control QD and EMMPF-QD

We first employed first-principles calculations to identify the adsorption sites and quantify the adsorption energies of [PF6]− and [EMIM]⁺ ions on the surface of CsPbBr₃ QDs. As illustrated in Fig. 1b, [PF6]− ions preferentially adsorb near the lead dangling bonds and cesium sites on the QD surface, exhibiting a strong adsorption energy of −279.8 kcal mol−¹ (Fig. 1d). In contrast, [EMIM]⁺ ions primarily bind to bromine dangling bonds (Fig. 1c) with an adsorption energy of −16.8 kcal mol−¹ (Fig. 1d). These results indicate that [PF6]− ions effectively passivates lead- and cesium-related defects, while [EMIM]⁺ ions targets bromine-related defects, collectively stabilizing the QD surface (Fig. 1a). For convenience, we denote QDs modified solely with [PF6]− and [EMIM]+ ions as PF-QDs and EMM-QDs, respectively, and those treated with the combined [EMIM]⁺/[PF₆]− ionic pair as EMMPF-QDs.

The Fourier transform infrared (FTIR) spectra confirmed the successful binding of multifunctional ligands to the QD surface: EMMPF-QDs exhibited characteristic phosphine–fluorine (P–F) bond (554 cm−¹) and carbon-nitrogen double bond (C=N) (1579 cm−¹) vibrations, which do not exist in control QDs (Fig. 1e), while PF-QDs and EMM-QDs showed only the P–F bond (559 cm−1) and the C=N bond (1565 cm−1) feature, respectively (Supplementary Fig. S1a, b). Furthermore, as shown in Supplementary Fig. S1c, compared with potassium hexafluorophosphate and 1-ethyl-3-methylimidazolium bromide, the peaks of P–F and C=N bonds in EMMPF-QDs had shifted, indicating that [PF6]− and [EMIM]+ ions interact with the dangling bonds of QDs. Similarly, as shown in Supplementary Fig. S2a, Raman peaks of potassium hexafluorophosphate and 1-ethyl-3-methylimidazolium bromide appeared at 788 cm−1 and 783 cm−1, respectively. Compared to them, the Raman peaks of PF-QDs (790 cm−1) and EMM-QDs (786 cm−1) both had shifted, which also confirms that [PF6]− and [EMIM]+ ions exist on the surface of QDs and interact with their dangling bonds. The Raman peak of EMMPF-QDs was also located at 786 cm−1, and had broadened, illustrating that it contains the information of [EMIM]+ and [PF6]− ions (Supplementary Fig. S2b). The X-ray diffraction (XRD) patterns of all QDs remained indexed to cubic phase CsPbBr₃ with no peak shifts upon ligand incorporation24,25, indicating that the ligands reside at the surface rather than within the lattice (Fig. 1f and Supplementary Fig. S3). The X-ray photoelectron spectroscopy (XPS) revealed systematic shifts of the binding energies of Pb 4f and Br 3d: [PF6]− ions with strong electronegativity lowered electron density and shifted energies upward26, while [EMIM]+ ions partially compensated by locally increasing electron density, underscoring their complementary roles in passivating dangling bonds and stabilizing the surface electronic structure (Fig. 1g and Supplementary Figs. S4 and S5).

We characterized the shape, size, and arrangement of QDs using a scanning transmission electron microscope (STEM). The shape of all QDs was quasi-cube, and their average sizes were not much different, which were all around 4.0 nm (Supplementary Fig. S6). STEM analysis showed that the control QDs were disordered and prone to metallic Pb formation due to strong interparticle coupling (the area circled by blue ovals) and the existence of lead dangling bonds (Fig. 2a). EMM-QDs had a higher degree of order compared to control QDs, but the interparticle coupling still exists (Fig. 2c; Supplementary Figs. S7a and S8a). Appropriate coupling is conducive to the improvement of the photoelectric performance of QD films, but overlarge coupling will increase the energy loss of QDs14,15,27. PF-QDs exhibited more uniform spacing and reduced Pb nanocrystals (Supplementary Fig. S7b), whereas EMMPF-QDs displayed the highest degree of order of 0.94 and resisted electron irradiation (Fig. 2a–d; Supplementary Figs. S7 and S8). The combined effects: P–F bonds filling Br vacancies to increase the spacing and suppress the coupling of QDs; imidazolium cations optimizing the degree of order of QDs and lowering charge-injection barriers of LEDs28,29,30.

Fig. 2: The shape, size, and arrangement of QDs and their corresponding optical properties.
Fig. 2: The shape, size, and arrangement of QDs and their corresponding optical properties.
Full size image

a, b The STEM images of a control QD and b EMMPF-QD. c, d The feature extraction and degree of order of c control QD and d EMMPF-QD. e, f The PL spectra of e control QD and f EMMPF-QD. g, h, The g PLQY and h time-resolved photoluminescence (TRPL) spectra of the control QD and EMMPF-QD films

These structural improvements translated into enhanced optical properties. The EMMPF-QDs retained strong quantum confinement yet exhibited a narrower photoluminescence (PL) spectrum (peak = 475 nm, FWHM = 26 nm) (Fig. 2e, f; Supplementary Figs. S9 and S10), and a higher PL quantum yield (QY) of 92% than the control QDs of 78%, with transient fluorescence lifetimes extended from 5.93 ns to 8.95 ns, indicating more efficient radiative recombination (Fig. 2g, h; Supplementary Figs. S11 and S12; Supplementary Table S1). In addition, we also characterized the defect density of QD films using electron-only devices. Compared to control QD, the trap-filling limit voltage (VTFL) of the device based on EMMPF-QDs decreased from 1.18 V to 0.51 V, showing a lower trap state density (Supplementary Fig. S13). The multifunctional molecule also improved stability: the EMMPF-QDs avoided the emission (~500 nm) related to defect-induced grain boundary diffusion observed in control QDs under a standard sun (1 sun) illumination (Supplementary Fig. S14) and retained 73% of initial PL intensity at 373 K after 5 min, better than the control QDs (Supplementary Fig. S15). The scanning electron microscope (SEM) confirmed near elimination of metallic Pb nanocrystals by introducing the multifunctional molecule (Supplementary Fig. S16). Importantly, the fishing net-like film morphology did not increase the leakage current of the LEDs (Supplementary Fig. S31).

Beyond surface stabilization, size reduction also lowered QD permittivity, weakening dielectric screening of the constraints of lattices on the excitons and intensifying Auger recombination at high luminance, thereby accelerating efficiency roll-off31,32. To counter this, we introduced the high permittivity imidazolium cation to suppress Auger recombination of QDs33,34,35. Excitation intensity-dependent PLQY confirmed this effect (Fig. 3a–d): at low excitation, the PLQYs trend reflected trap-state densities (control QD > PF-QD > EMM-QD > EMMPF-QD), whereas at higher intensities, the EMMPF-QD film reached the highest PLQY, consistent with reduced Auger losses via enhanced dielectric screening. Transient absorption spectroscopy corroborated this behavior (Fig. 3e–h): the ΔA peak of the control QDs exhibited the largest decay under the irradiation of a high flux laser of 26 μJ cm−2, the PF-QDs showed similar dynamics, while the ΔA peak of the EMM-QDs and EMMPF-QDs remained nearly constant, confirming the inhibitory effect of [EMIM]+ ions on Auger recombination. Furthermore, by using temperature-dependent PL spectra, we found that the PL spectral area of EMMPF-QD decreased the fastest with the increase in temperature (Supplementary Fig. S17), and its exciton binding energy was the smallest (Supplementary Fig. S18), which indicates that the Coulombic force between electrons and holes in EMMPF-QD is the weakest. This confirms that [EMIM]+ ions enhance the dielectric screening effect of QDs by increasing their permittivity, thereby weakening the Coulombic force of electrons and holes in the QDs and reducing their exciton binding energy, achieving the purpose of suppressing Auger recombination.

Fig. 3: Inhibition of Auger recombination of QDs by the multifunctional molecule.
Fig. 3: Inhibition of Auger recombination of QDs by the multifunctional molecule.
Full size image

a–d The excitation intensity-dependent PLQY of a control QD, b PF-QD, c EMM-QD, and d EMMPF-QD. e–h Time-dependent transient absorption spectra of e control QD, f PF-QD, g EMM-QD, and h EMMPF-QD

To assess the impact of [EMIM]+ ions on efficiency roll-off, we fabricated the perovskite LEDs based on control QDs and PF-QDs using the structure in Supplementary Fig. S19, and the interfaces between each layer of the LED were relatively clear and smooth. Ultraviolet photoelectron spectroscopy (UPS) determined the secondary electron cut-off edge and Fermi edge (Supplementary Fig. S20), and calculated the valence-band maximum (VBM) of QDs, revealing that [PF6]− ions increased the energy barrier between QDs and Poly[bis(4-phenyl)(2,4,6-triMethylphenyl)aMine] (PTAA) (Supplementary Fig. S21). Despite this, the PF-QD-based LED exhibited a turn-on voltage of 3.2 V, lower than the LED based on the control QDs of 3.6 V, and enhanced current density, suggesting improved conductivity and reduced trap states (Supplementary Fig. S22). Furthermore, its peak luminance and EQE doubled compared to that of the control QD-based LED (5176 cd m−², 18.52% vs. 1735 cd m−², 8.39%). However, its efficiency roll-off remained substantial: when the luminance reached 1828 cd m−2, EQE had dropped to 93% of peak (Supplementary Fig. S23). Generally speaking, the LED devices applied to HD display panels need to maintain a high EQE at a luminance of at least 2000 cd m−2. In some special application scenarios (automotive displays, outdoor displays), it is even necessary for LEDs to still be able to work efficiently at a brightness of 4000 cd m−2. Moreover, their color purity was insufficient for HD displays (CIEy > 0.1; Supplementary Fig. S24).

In contrast, the VBM of EMM-QDs and EMMPF-QDs shifted upward, eliminating the charge injection barrier with PTAA (Fig. 4a; Supplementary Figs. S25 and S26), thereby enabling balanced charge injection and reducing charge accumulation under high current densities. Compared with the control QD-based LED, the EL intensity of the EMMPF-QD-based LED had almost no attenuation, which means that the latter has more balanced charge injection and less interfacial accumulation (Supplementary Fig. S27). For the control QD-based LED, the charge accumulation caused by inconsistent electron and hole injection or transport rates will gradually become prominent under high current density, leading to an excess of one carrier type accumulating at interfaces, increasing the chances of non-geminate recombination or simply leakage current without forming excitons, which will exacerbate the efficiency roll-off. Consequently, the EMM-QD- and EMMPF-QD-based LEDs exhibited a low turn-on voltage (Von) of 2.8 V and 2.6 V, respectively, and achieved a peak luminance (Lmax) of 11,470 cd m−² and 13,389 cd m−², respectively, better than the control QD- and PF-QD-based LEDs (Fig. 4b and Supplementary Fig. S28). As shown in Fig. 4c and Supplementary Fig. S29, suppression of Auger recombination by imidazolium cations yielded an exceptionally low efficiency roll-off; among them, the EMMPF-QD-based LED had a peak EQE of 20.02% at 6441 cd m−², and the EQE remains 18.47% at 9587 cd m−², also superior to the control QD- and PF-QD-based LEDs. And within the luminance range of 1920 cd m−2 to 13,389 cd m−2, the EQEs of the EMMPF-QD-based LED always remained above 15%, which has expanded the application scenarios of LED in the field of HD display (Table 1)5,7,9,12,13,17,18,19,36,37. Its EL emission at 472 nm showed enhanced color purity (CIEy = 0.091), meeting Rec. 2020 standards for HD display applications (Fig. 4d, e). Furthermore, the LED based on EMMPF-QDs was a Lambertian emitter, and its EL intensity in all directions was nearly equal (Supplementary Fig. S32). Significantly, the EL spectra of the devices based on PF-QDs, EMM-QDs, and EMMPF-QDs had undergone a slight blue shift compared to their PL spectra (Figs. 2f and 4d; Supplementary Figs. S9, S24, and S30). The possible reason for this result is that under an electric field, [PF6]− and [EMIM]+ ions can slightly broaden the band gap of QDs.

Fig. 4: The performance of the EMMPF-QD-based LEDs.
Fig. 4: The performance of the EMMPF-QD-based LEDs.
Full size image

a The structure and energy level diagram, b the J–L–V curve, c the EQE-L curve, the illustration is the photo of the device in operation, d the EL spectrum, and e the CIE coordinates. f The statistical diagram of EQE peak and corresponding luminance of the reported blue perovskite LEDs. g The T50 of the LED at a constant current of 1.5 mA cm−2. h The statistical diagram of T50 and the corresponding initial luminance of the reported blue perovskite LEDs. i The variation of CIE coordinates

Table 1 The EQE corresponding to different luminance of the reported blue perovskite LEDs (EQE peak ≥ 12%, CIEy < 0.1)
Full size table

Benchmarking against literature, our device achieved the highest reported value of peak EQE × corresponding luminance for blue perovskite LEDs (CIEy < 0.1, EQE ≥ 12%), exceeding the previous best by 3.6-fold (Fig. 4f)5,7,9,13,17,37. Thermal imaging showed the EMMPF-QD-based LEDs operated at a significantly reduced temperature (33.0 °C vs. 47.6 °C for control), confirming the multifunctional molecule effectively mitigated device self-heating caused by Auger recombination (Supplementary Fig. S33). The elevated temperatures can increase non-radiative decay rates (thermal quenching) and cause degradation of materials or interfaces, which is also the main cause for the efficiency roll-off. The reproducibility of LEDs was high, with an average EQE of 16.26 ± 0.02% across 15 devices, which is significantly higher than the control QD-based LEDs (Supplementary Fig. S34).

Operational stability, a major challenge in high-performance blue perovskite LEDs, was significantly improved. While prior reports with high EQE and color purity typically suffer from short lifetimes (T₅₀ ~144 min at 20 cd m−2), our EMMPF-QD-based LED achieved a T₅₀ of 692 min at an initial brightness of 106 cd m−² (Fig. 4g), which is 24-fold that of the LED based on control QDs (Supplementary Fig. S35), and its initial luminance × T50 was 25 times greater than the best previously reported device, which is prepared by vacuum thermal evaporation process (Fig. 4h)5,7,9,13,17. Moreover, after operation for 12 h, the input voltage of the EMMPF-QD-based LED only increased from 3.8 V to 5.1 V; in contrast, the input voltage of the LED based on control QDs rapidly increased from 4.9 V to 7.1 V after working for 0.5 h. This indicates that the former has a slower increase in device resistance and a more stable interface during operation (Supplementary Fig. S36). The EL spectra of the LED based on EMMPF-QDs compared to the control QD-based LED remained stable during voltage ramps and long-term operation (Fig. 4i; Supplementary Figs. S37 and S38), underscoring the robustness conferred by multifunctional molecule passivation. These advances represent a major step toward practical blue perovskite LEDs for HD display technologies.

Discussion

We demonstrate that a multifunctional molecule effectively stabilizes perovskite QD surfaces, simultaneously addressing carrier trapping, interdot coupling, and Auger recombination—key challenges limiting blue perovskite QLED performance. The synergistic action of [PF6]− anions and [EMIM]+ cations yields QDs with enhanced photoluminescence efficiency, prolonged carrier lifetimes, and superior thermal and illumination stability. These advances enable blue perovskite QLEDs that combine high color purity (CIEy = 0.091) with outstanding peak EQE (20.02%) at high luminance (6441 cd m−²) and exhibit markedly suppressed efficiency roll-off. Importantly, the devices also achieve extended operational lifetimes with stable emission spectra throughout. Our findings provide a promising pathway toward robust, high-performance blue perovskite QLEDs suitable for demanding display applications.

Materials and methods

Materials

Octadecene (ODE) was purchased from Innochem. We bought Cesium carbonate (Cs2CO3) and oleic acid (OA) from Alfa Aesar. We bought Oleylamine (OAm) from Meryer. Lead bromide (PbBr2) was purchased from Youxuan. We bought Zinc bromide (ZnBr2), methyl acetate (MA), hexane, rubidium bromide (RbBr), potassium hexafluorophosphate, 1-ethyl-3-methylimidazolium bromide, EMIMPF₆, lithium fluoride (LiF) and Aluminum (Al) from Aladdin. Octane was purchased from Macklin. We bought phenethylammonium bromide (PEABr), 2,4,6-Tris[3-(diphenylphosphinyl)phenyl]-1,3,5-triazine (PO-T2T), 1,3,5-Tris(N-phenylbenzimidazol-2-yl)benzene (TPBi), PTAA, and poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT: PSS) from Yuri Solar. All reagents did not require further purification.

Preparation of CsPbBr3 QDs

We added ‌100mg Cs2CO3‌, 0.7mL OA‌, and 10mL ODE to a three-neck flask, then stirred at 40, 60, and 100 °C for 10 min each under vacuum until it became clear and transparent. Finally, the solution was kept at ‌100 °C under a nitrogen atmosphere‌. The above is the preparation of cesium precursors. Meanwhile, 0.376 mmol PbBr2‌, 3 mmol ZnBr2, and 10 mL ODE‌ were introduced in a three-neck flask, and sequentially agitated at ‌40, 60, 100, 120 °C‌ for ‌10 min each‌ under vacuum. Under a nitrogen atmosphere, ‌4.5 mL OA‌ and ‌4.5 mL OAm‌ were infused into the flask. The temperature was set to ‌115–120 °C‌. The above is the preparation of lead precursors. And 1.6 mL cesium precursors were rapidly infused into the lead precursor solution. After ‌5 s, the three-neck flask was transferred to an ‌ice bath‌ to halt the reaction38.

Purification of CsPbBr3 QDs

The stock solution was diluted with ‌50 to 80 mL of MA‌ and centrifuged at ‌10,000 rpm for 1 min‌ to precipitate the QDs, which were redispersed in ‌1 mL octane‌. As shown in Supplementary Fig. S39, added 2 ~ 4 mL of MA to 1 mL of the QD octane solution, and the mixture was centrifuged at 12,000 rpm for 1 min to precipitate the QDs. Resuspend the precipitated QDs in 0.5 mL of hexane for further use38.

Passivation of CsPbBr3 QDs

In total, 79 mg RbBr‌ was added to the purified QD dispersion. The mixture was stirred for ‌90 min‌ at ambient temperature38, and after filtering to obtain the control QDs (Supplementary Fig. S39). Totally, 2.5–5 mg ‌EMIMPF₆‌ were added to the control QD dispersion and stirred for ‌10–60 s to obtain the EMMPF-QD dispersion by ‌filtration. Among them, the EMMPF-QDs obtained by stirring for 30 s had the best optical performance and device performance (Supplementary Figs. S40 and S41). Similarly, 2.5–5 mg of potassium hexafluorophosphate was added to the control QDs, and the PF-QD dispersion was obtained by ‌filtration; 2.5–5 mg of 1-ethyl-3-methylimidazolium bromide was added to the control QDs, and the EMM-QD dispersion was obtained by ‌filtration. The filtered QD dispersion can be ‌directly applied‌ to LED fabrication without additional processing.

Characterization of CsPbBr3 QDs

A ‌Bruker D8 Discover diffractometer‌ was used to gauge the XRD patterns‌. The FTIR spectra‌ were performed on a Bruker ALPHA II spectrometer. A HORIBA LabRAM Odyssey spectrometer was used to measure the Raman spectra. The XPS‌ was performed on a ‌Thermo Fisher Escalab Xi+ spectrometer‌. A ‌Shimadzu UV-2600 spectrophotometer was used to document the UV–visible absorption spectra. ‌A ‌Hitachi F-4600 fluorescence spectrometer‌ was used to acquire the PL spectra. The ‌STEM images (at ‌200 kV acceleration voltage‌) were obtained using a ‌Thermo Fisher Talos F200X‌ electron microscope. A ‌Zeiss GeminiSEM360‌ electron microscope was used to capture the SEM‌ images. The secondary electron cut-off edge and Fermi edge were acquired by a Thermo Fisher Escalab 250Xi spectrometer, and the VBM of QDs was calculated by the following formula:

$${\text{Work}}\,{\text{function}}=21.22\,{\text{eV}}-{\text{cut}}\hbox{-}{\text{off}}\,{\text{edge}}$$
(1)
$$\mathrm{VBM}=-\mathrm{work}\,\mathrm{function}-\mathrm{Fermi}\,\mathrm{edge}$$
(2)

The light source used to test the illumination stability of QD films was a Newport Oriel Sol3A solar simulator. PLQYs, TRPL spectra, and excitation intensity-dependent PLQYs were recorded on an Edinburgh FLS1000. A HORIBA Fluorolog-3 fluorescence spectrometer was used to collect the temperature-dependent PL spectra. The femtosecond TA spectra were acquired using a ‌TA-ONE-1 spectrometer‌ with an excitation wavelength of ‌400 nm‌, enabling ultrafast dynamics studies with high temporal resolution‌38. The structure of the electron-only devices was ITO/ZnO/QDs/TPBi/LiF/Al. We used the Gabor filter to extract features from the STEM images of the QDs, and processed these features using the Gaussian distribution fitting and the second moment of Chebyshev polynomial expansion to quantify the degree of order of the QDs39.

Computational details

The adsorption sites and adsorption energies of [PF₆]− and [EMIM]⁺ ions on the QD surface were calculated by the Focite module in Materials Studio. The convergence threshold: 2 × 10−5 kcal mol−1 in energy, 1.2 × 10−8 kcal mol-1 Å−1 in force, 1 × 10−5 Å in displacement, and the forcefield is universal38.

Fabrication and measurement of LEDs

The indium tin oxide (ITO) based glasses were washed with ITO cleaning solution and ultrasonically cleaned sequentially in deionized water and ethanol for 30 min each. Then dried thoroughly under a stream of pure nitrogen gas, and treated with UV-ozone cleaning for 20 min. The PEDOT:PSS solution was static spin-coating on the ITO substrate at 4000 rpm for 40 s. Then annealed at 150 °C for 10 min in ambient air. After that, the 5 mg mL−1 PTAA chlorobenzene solution was dynamic spin-coating on the PEDOT:PSS film at 4000 rpm for 45 s, then annealed at 150 °C for 15 min in a nitrogen glove box. Then, the PEABr solution (10 mg mL−1) was dynamic spin-coating on the PTAA film, and annealed at 80 °C for 10 min. The blue CsPbBr3 QD octane dispersion was dynamic spin-coating on the PEABr at 2000 rpm for 40 s. Whereafter, the ITO based glasses were transferred to the vacuum thermal evaporator, and 50 nm of TPBi/PO-T2T, 1 nm of LiF, and 100 nm of Al were successively deposited on the QD film (<3 × 10−4 Pa). The properties of unencapsulated perovskite QLEDs were measured via a CS-3000 spectroradiometer and a Keithley 2400 source meter in the nitrogen glove box38. The temperature of perovskite QLEDs during operation was recorded by a Uti320e infrared thermal imaging camera. The EL intensities of LEDs in different directions were measured by HAMAMATSU C9920-11. The transient EL spectra of the LEDs use a Keysight 81150 A as the pulser.