Mesoscale ordered assembly of Er³⁺-doped quantum dots enables efficient 1.55 µm electroluminescence

Abstract

Efficient, electrically driven 1.55 µm light sources are highly desirable for telecom-band emission, on-chip optical sensing, and integrated near-infrared photonics, yet their realization remains a grand challenge. Direct integration of Er³⁺ into semiconducting perovskite quantum dots (QDs) offers superior potential by enabling intrinsic charge transport, but this pathway is stifled by kinetic barriers that yield polydisperse nanocrystals and random packing, collectively inducing leakage and quenching. Here, we report mesoscale ordering of Er³⁺-doped CsPbCl₃ QDs with efficient 1.55 µm electroluminescence by coupling kinetically controlled synthesis with hydrogen-bonding-directed assembly. Using myristoyl chloride as a slow-release chloride source, we synthesize monodisperse QDs while in-situ generating amide ligands that form interdot N–H···O = C hydrogen-bonding networks. These networks direct assembly into mesoscale ordered assemblies with preferential {100} face-to-face orientation. The resulting architecture yields a record EQE of 3.75% and maximum radiance of 323.31 mW sr⁻¹ m⁻², ~10 times higher than the disordered controls, with an operational stability of 197 minutes (~7-fold improvement). This establishes hierarchically ordered semiconductor assemblies as promising candidate for telecom-band optoelectronics.

Efficient 1.55 micrometer light sources are needed for telecom and infrared photonics, but realizing them with perovskite quantum dots is challenging due to bandgap limitations. Li et al. used Er3 + -doping and hydrogen-bonded ordered assembly in CsPbCl3 to enable efficient electroluminescence.

Introduction

Efficient and stable near-infrared electroluminescent sources are central components for on-chip optical interconnects^1–7, spectroscopic sensing and biomedical imaging^8–10. Lanthanide ions exhibit narrow, atomically resolved 4f–4 f transitions^11,12, millisecond-scale excited-state lifetimes^13,14 and excellent photostability^15,16, making them attractive emitters in the near-infrared region^9,17,18. In particular, the Er³⁺ 1.55 µm transition coincides with the minimum-loss window of silica optical fiber and therefore lies within the telecom C-band, making it attractive for integrated near-infrared photonic applications¹⁹. To date, III–V lasers dominate this spectral region, but their integration with silicon photonics remains costly and technologically complex^20,21.

Solution-processed NIR QD-LEDs based on PbS have achieved EQEs of up to 11.8% at 1550 nm²², while InAs-based devices have reached 20.5% at ~900 nm but remain limited at telecom C-band wavelengths²³. Both systems, however, suffer from broad emission linewidths (>100 nm) arising from size polydispersity. In contrast, lanthanide-based emitters offer atomically defined, ultra-narrow 4f–4 f transitions (FWHM < 20 nm) with exceptional photostability, representing a fundamentally distinct approach to telecom-band electroluminescence. Solution-processable, electrically driven lanthanide emitters would therefore provide a more accessible pathway to electrically driven, telecom C-band emitters compatible with integrated photonic platforms.

Direct integration of lanthanides into semiconducting quantum dot (QD) matrices, particularly metal-halide perovskites, offers a promising route to electrically driven telecom-band emission by combining charge transport with broadband optical sensitization^24–26. Progress is nevertheless impeded by two coupled synthesis and assembly challenges. At the nanocrystal scale, achieving efficient Er³⁺ incorporation requires elevated synthesis temperatures^27,28 ( ≥ 240 °C) to surmount cation-exchange kinetic barriers, a condition that promotes uncontrolled nucleation, broad size distributions and elevated trap-state densities²⁹. At the film scale, random QDs packing creates percolative leakage pathways and trap-assisted tunneling currents; under electrical bias, these effects engender local charge accumulation, excessive Joule heating, and exciton quenching of the long-lived Er³⁺ excited state, resulting in severe efficiency roll-off at practical radiance levels.

Here, we employ a dual-level hierarchical design strategy combining kinetically controlled nanocrystal synthesis with ordered assembly engineering. Using myristoyl chloride as a programmable slow-release chloride source, we synthesize monodisperse, low-defect Er³⁺-doped CsPbCl₃ QDs while simultaneously generating amide ligands in situ. The amide moieties establish interdot N–H···O = C hydrogen-bonding networks during film formation, directing spontaneous assembly into mesoscale ordered assemblies with preferential {100} face-to-face orientation. The ordered architecture optimizes charge pathways and suppresses exciton quenching; devices fabricated from these films exhibit 1.55 µm electroluminescence with a peak EQE of 3.75%, a maximum radiance of 323.31 mW sr⁻¹ m⁻², and markedly improved operating stability. Our results demonstrate that multiscale structural control—from monodisperse nanocrystal synthesis to hydrogen-bonded ordered assembly—is essential for translating lanthanide photophysics into high-brightness, electrically driven devices and provides a materials blueprint for telecom-band, solution-processed light sources.

Results

We synthesized Er³⁺-doped CsPbCl₃ QDs and optimized their nucleation and growth kinetics to achieve near infrared emission in the 1.55 µm (Fig. 1). In conventional Lanthanide-doped perovskite QDs synthesis using trimethylsilyl chloride (TMS–Cl) as the chloride source^30–32, precursor conversion crosses the nucleation threshold within an extremely short time window, followed by rapid monomer depletion. When the system approaches the solubility line, growth enters a monomer-starved regime where particle sizes become mismatched and broadened³³ (Fig. 1a, stages I–III). To reverse this kinetic trajectory, we introduce myristoyl chloride (Myr–Cl), a slow-reacting acyl chloride that maintains high and stable monomer activity throughout the reaction. This shifts crystal growth from reaction-limited to surface-diffusion-limited, inherently accompanying size focusing and yielding highly monodisperse CsPbCl₃:Er³⁺ QDs (Fig. 1b).

Fig. 1. Kinetic control and H-bond–guided ordering of Er³⁺-doped CsPbCl₃ QDs.

a Conventional chloride precursors (e.g., trimethylsilyl chloride, TMS–Cl) trigger a brief nucleation burst followed by rapid monomer depletion (regions I–III), pushing growth below the solubility line and causing size broadening. b Using myristoyl chloride (Myr–Cl) slows precursor conversion, sustains monomer activity into region III, and drives size focusing. c During synthesis Myr–Cl undergoes in situ aminolysis to generate amide ligands that form inter-dot N–H···O = C hydrogen bonds, bridging neighboring nanocubes and guiding face-to-face packing into ordered films, which in turn improves device performance (higher external quantum efficiency, EQE, and slower roll-off). QDs quantum dots.

Beyond kinetic control in solution, the chemistry of Myr–Cl also programs the solid-state assembly. During synthesis, Myr–Cl undergoes aminolysis with alkylamines to generate amide ligands in situ. Unlike oleate capping, the –CONH– headgroups establish directional N–H···O = C hydrogen bonds between neighboring QD surfaces, thereby strengthening and orienting interdot coupling through casting. During spin-coating and antisolvent quenching, these H-bond networks bias facet-matched, face-to-face stacking and layer-by-layer compaction, narrowing interdot-spacing dispersion and establishing ordered assembly (Fig. 1c). This ligand-programmed, hierarchical pathway—kinetic size focusing in solution followed by H-bond–guided assembly in films—reduces leakage current, balances carrier injection (higher peak EQE), and markedly suppresses efficiency roll-off under high drive^34,35.

Kinetically controlled synthesis yields monodisperse Er³⁺-doped QDs

To reveal the chemical origin of this kinetic modulation, we analyzed the reaction intermediates. Spectroscopic evidence confirms that Myr–Cl undergoes in situ acyl–amine exchange to generate branched amide-type ligands^36,37 (Supplementary Fig. 2). Compared with TMS–Cl, these ligands exhibit enhanced surface binding (Supplementary Fig. 3) and possess bulky alkyl chains that impose steric hindrance at the nanocrystal interface^38,39. This effectively slows down surface atom diffusion and stabilizes the growth front, contributing to the improved size focusing (Fig. 2a, b).

Fig. 2. Synthesis and structural characterization of monodisperse Er³⁺-doped CsPbCl₃ QDs.

a, b Schematics of the traditional fast-growth route (a) and the present slow-growth, amide-ligand strategy (b). c Transmission electron microscopy (TEM) images of the control (TMS–Cl, top) and target (Myr–Cl, bottom; 13.1 ± 0.8 nm) QDs. Scale bars, 100 nm. d High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of a single QD along the [100] zone axis, edge and center regions boxed. Scale bar, 2 nm. e Magnified HAADF-STEM images of the edge (top) and center (bottom) with atomic models and fast Fourier transforms (FFTs).f Density functional theory (DFT) formation energies for Er at the Cs site, Pb site, interstitial gap and Pb-and-gap configurations g Schematic of the Pb→Er substitution within the [PbCl₆]^4- octahedron. h Atomic-resolution HAADF image of the target QD. Scale bar, 1 nm. i The same region overlaid with atomic-level elemental maps of Er (green) and Pb (red), showing complete spatial overlap and confirming uniform substitutional doping. Scale bar, 1 nm.

We used transmission electron microscopy (TEM) and high-angle annular dark-field (HAADF) imaging to examine the morphology and size distribution of the resulting QDs. The CsPbCl₃: Er³⁺ QDs synthesized with Myr–Cl exhibit highly uniform cubic morphology with an average diameter of 13.1 ± 0.8 nm, whereas those prepared with TMS–Cl show irregular shapes and a broader size distribution ranging from 12 to 20 nm (Fig. 2c, Supplementary Fig. 4). Low- and high-magnification TEM and HAADF (Supplementary Fig. 4) images further reveal densely packed, monodisperse nanocubes without noticeable aggregation, indicating well-controlled nucleation and growth kinetics under the Myr–Cl–mediated synthesis.

We further used aberration-corrected HAADF-STEM to directly visualize the atomic structure of CsPbCl₃:Er³⁺ QDs (Fig. 2d, e). The images show a perfectly ordered cubic lattice without detectable secondary phases, indicating that Er incorporation does not disrupt the perovskite framework. Density-functional-theory (DFT) calculations reveal that substitution of Pb²⁺ sites by Er³⁺ is the most energetically favorable configuration, whereas Cs-site or interstitial incorporation is thermodynamically unstable (Fig. 2f, g, Supplementary Fig. 5). Atomic-resolution EDS mapping (Fig. 2h, i, Supplementary Fig. 6) further confirms this substitutional model, showing complete spatial overlap between Pb and Er signals, evidencing uniform distribution of Er³⁺ ions within the lattice rather than surface segregation. The XRD (Supplementary Fig. 7) peaks exhibit a slight shift toward higher diffraction angles, consistent with lattice contraction caused by Er³⁺ replacing Pb²⁺, further validating the substitutional doping mechanism.

We then used optical spectroscopy to evaluate how kinetic regulation and uniform doping affect the structural and electronic quality of the QDs. Compared with TMS–Cl–derived QDs, the Myr–Cl samples exhibit a sharper excitonic absorption peak and a reduced Urbach tail, indicating narrower size distribution and lower trap density. Correspondingly, the photoluminescence spectra show a markedly narrower full width at half maximum (FWHM = 10.4 nm) than that of TMS–Cl–based QDs (23.3 nm), reflecting improved monodispersed and phase purity (Supplementary Fig. 8). The narrower size distribution and reduced defect density suppress trap-assisted carrier loss in the emissive layer, while the higher monodispersity also provides a more favorable basis for compact and ordered film packing during assembly. As a result, the benefits of the slow-release halide chemistry propagate from nanocrystal synthesis to film organization, ultimately contributing to reduced leakage, more uniform carrier transport, and improved electroluminescent performance. Beyond the band-edge emission at ~410 nm, the CsPbCl₃:Er³⁺ QDs exhibit characteristic Er³⁺ NIR emissions at 985 nm and 1550 nm (Supplementary Fig. 9). ICP-OES analysis identifies the optimal Er incorporation at 6.99 at.%, yielding a peak PLQY of 20.47% (Supplementary Fig. 10 and Supplementary Table 1).

Hydrogen-bonding networks direct QD assembly into ordered films

Having achieved monodisperse QDs, we investigated how the amide ligands influence collective film organization. We hypothesized that the amide groups (–CO–NH–) could form interdot hydrogen bonds, potentially directing assembly into ordered structures. To test this, we performed temperature-dependent FTIR spectroscopy on quantum dot films. We observed that the N–H stretching band at 3000–3400 cm⁻¹ broadened and shifted to lower frequency with increasing temperature (from room temperature to 60 °C) (Supplementary Fig. 11), characteristic of hydrogen-bond weakening under thermal excitation. This temperature-dependent behavior confirmed that N–H···O = C hydrogen bonds persist in the solid state and mediate interdot interactions⁴⁰.

We then employed grazing-incidence small-angle X-ray scattering (GISAXS) (Fig. 3a–c) and grazing-incidence wide-angle X-ray scattering (GIWAXS) (Fig. 3d-f) to directly visualize film structure across multiple length scales. We prepared control films using TMS–Cl as the chloride source and target films using Myr–Cl; both were measured to be approximately 35 nm thick with similar composition, providing a fair basis for comparing diffraction behavior. GISAXS patterns from target films (Fig. 3b) showed sharp, well-defined diffraction spots in both in-plane (q_xy) and out-of-plane (q_z) directions, indicating three-dimensional mesoscale ordering of QDs. In contrast, control films (Fig. 3a) exhibited only diffuse, isotropic scattering rings characteristic of randomly oriented assemblies. Radial integration of GISAXS patterns revealed distinct first-order peaks at q = 0.053 Å⁻¹ for target films and q = 0.047 Å⁻¹ for control films (Fig. 3c), corresponding to d-spacings of 11.8 nm and 13.4 nm, respectively⁴¹. The sharper peak profile in target films indicates closer packing and mesoscale ordered assembly mediated by hydrogen-bonding interactions. The dominant in-plane GISAXS reflection is assigned to the (110) plane, giving α_SL = d₁₁₀$\sqrt{2}$ = 16.7 nm and an estimated interdot gap of ~3.6 nm. The structural coherence lengths extracted from the small-q peak widths are ~55.6 nm (target) and ~33.1 nm (control), confirming the larger ordered-domain size in the target film.

Fig. 3. Mesoscale ordered assembly and crystallographic alignment in amide-ligand films.

a, b Grazing-incidence small-angle X-ray scattering (GISAXS) maps of the disordered (a) and ordered (b) films. c Small-q radial line-cuts of the ordered (blue, top) and disordered (red, bottom) films, revealing first-order peaks at q = 0.053 Å⁻¹ (Δq = 0.0113 Å⁻¹) and q = 0.047 Å⁻¹ (Δq = 0.0190 Å⁻¹), corresponding to d-spacings of ~11.8 and ~13.4 nm and structural coherence lengths of ~55.6 and ~33.1 nm, respectively. d, e Grazing-incidence wide-angle X-ray scattering (GIWAXS) patterns of the disordered (d) and ordered (e) films. f Azimuthal profiles along the {100} ring for the ordered (blue) and disordered (red) films. g, h Schematics of the cubic perovskite (100) plane (g) and of the nanocube supercell highlighting the 45° relationship between the {110} and {100} planes (h). i Schematic comparing the mesoscale-ordered, {100}-oriented assembly of the ordered film (top) with the disordered arrangement of the control film (bottom). In a, b, d and e, the colour scale represents scattering intensity.

Despite both films having similar thickness, GIWAXS patterns differed dramatically (Fig. 3d, e): target films exhibited an approximately three-fold greater diffraction intensity than controls (Fig. 3e, Supplementary Fig. 12), suggesting QDs within the film are arranged with enhanced orientational order³⁴. To further determine the preferred orientation of QDs, we performed azimuthal integration of diffraction peaks^42,43. The (100) reflection from target films displayed a sharp azimuthal peak centered at 0° ~ 10° (Fig. 3f), indicating that the cubic perovskite {100} planes are preferentially aligned parallel to the substrate. Control films, by contrast, showed nearly flat azimuthal profiles consistent with random orientations.

We further analyzed the (110) reflection in azimuthal space (Supplementary Fig. 12), which revealed a characteristic peak at approximately 45° relative to the substrate normal—exactly the expected angle between {110} and {100} planes in a cubic lattice (Figs. 3g and 3h). This angular relationship confirms that QDs adopt a well-defined orientational registry: the {100} crystal faces align parallel to the substrate, while {110} planes tilt at 45° (Fig. 3i, top panel). In disordered control films, QDs pack randomly without preferential orientation (Fig. 3i, bottom panel). Together with the GISAXS spacings, these data establish that the in-situ amide ligands drive face-to-face alignment and mesoscale ordered assembly, plausibly via interdot N–H···O = C networks that favor facet-matched, hydrogen-bond–rich configurations during film formation. This interpretation is further supported by a dosage-dependent ligand-shell analysis, which shows that the solid-state QD surface is not a fully converted amide shell but a mixed OA/OAm/amide ligand shell, and that the most ordered packing is obtained at an intermediate ligand composition where directional amide-mediated interactions and long-chain steric effects are best balanced (Supplementary Fig. 13and Supplementary Table 2).

Energy transfer from CsPbCl₃ matrix efficiently sensitizes Er³⁺ emission

We investigated the excited-state dynamics and energy-transfer processes in CsPbCl₃:Er³⁺ QDs using transient absorption (TA) and time-resolved photoluminescence (TRPL) spectroscopy⁴⁴ (Fig. 4, Supplementary Figs. 14, 15). Compared with pristine CsPbCl₃ (Fig. 4a), the Er³⁺-doped samples (Fig. 4b) exhibit a faster bleach recovery of the host exciton, decreasing from ~365 ps to ~100 ps, indicating energy transfer from the QD matrix exciton to Er³⁺. The TRPL maps further reveal a sequential emission process (Fig. 4c, d): the band-edge emission of CsPbCl₃ decays within a nanosecond window, while Er³⁺ near-infrared bands rise on the millisecond scale with distinct onset times—the 985 nm band appears at ~0.5 ms and the 1550 nm band builds up at ~1.5–3.2 ms after nonradiative relaxation to ⁴I_13/2. The time-gated PL spectra (Fig. 4e) confirm that Er³⁺ excited states are populated stepwise: energy harvested by the CsPbCl₃ matrix is first funneled into Er³⁺(⁴I_11/2) (yielding 985 nm emission), followed by multiphoton relaxation to ⁴I_13/2 and subsequent 1550 nm emission.

Fig. 4. Matrix-to-Er³⁺ energy transfer and sequential NIR emission.

a, b Femtosecond transient-absorption spectra of pristine CsPbCl₃ and CsPbCl₃:Er³⁺. The exciton bleach recovers faster upon Er³⁺ doping (~365 ps → ~100 ps), indicating matrix→Er³⁺ energy transfer. c, d Time-resolved PL maps: the host band-edge emission decays within <0.5 µs, while Er³⁺ NIR bands rise on the millisecond scale. Colour scale, emission intensity. e Time-gated spectra show stepwise population—first ⁴I_11/2 (985 nm), then ⁴I_13/2 (1550 nm). f Excitation spectra monitored at 985 and 1550 nm track the CsPbCl₃ absorption, confirming broadband host sensitization. g Mechanistic illustration: photoexcited QD → ⁴I_11/2 → ⁴I_13/2 → ⁴I_15/2, yielding sequential 985 and 1550 nm emissions.

To verify broadband host sensitization and identify the preferred injection channel, we recorded excitation spectra at 985 nm and 1550 nm, both of which closely follow the steady-state absorption of CsPbCl₃ (Fig. 4f). This indicates that the perovskite could harvest visible photons and sensitize Er³⁺ emission. The host-to-Er³⁺ energy-transfer efficiency, estimated from the TA lifetime shortening (Supplementary Fig. 15), is ~70.2%, confirming efficient matrix-to-Er³⁺ sensitization. The mechanism is summarized in Fig. 4g as matrix exciton →⁴I_11/2 → ⁴I_13/2 → ⁴I_15/2; the ⁴I_11/2 branch triggers the ~0.5 ms 985 nm emission, while population transfer to ⁴I_13/2 establishes the 1550 nm emission at ~1.5–3.2 ms. Notably, the ns-lifetime of the matrix and the ms-lifetime of Er³⁺are naturally separated in time, providing temporal evidence for sequential population and indicating that the lanthanide long-lived emission is not significantly quenched, consistent with a low-defect environment and efficient energy transfer.

Device performance characterization

Encouraged by the improved film order and monodisperse quantum dot synthesis, we fabricated LEDs to evaluate electroluminescence performance (Fig. 5). We used a device architecture consisting of ITO (anode): PEDOT: PSS (hole-injection layer); poly-TPD (hole-transport layer); CsPbCl₃:Er³⁺ QDs (active layer); PO-T2T (electron-transport layer); Liq; and Al (cathode) (Fig. 5a). Cross-sectional scanning electron microscopy (SEM) image shows each corresponding layer (Supplementary Fig. 16). The energy level alignment measured by ultraviolet photoelectron spectroscopy enables efficient charge injection from both electrodes (Fig. 5b, Supplementary Fig. 17). The voltage-dependent EL spectra covering both the visible and NIR regions are provided in Supplementary Fig. 18.

Fig. 5. Device architecture and performance enabled by ordered Er³⁺-doped QD films.

a Schematic of the device structure. b Energy-level diagram showing balanced carrier injection. c Current density–voltage (J–V) and radiance–voltage (R–V) characteristics of the target (blue) and control (brown) devices; the target reaches a radiance of 323.31 mW sr⁻¹ m^-2, versus 33.86 mW sr⁻¹ m^-2 for the control. d External quantum efficiency (EQE) versus voltage: 3.75% (target) versus 0.38% (control). e Operational stability at 10 mA cm⁻²: the target shows T₅₀ = 197 min, far exceeding the control (28 min). f Normalized EQE retention versus radiance; the target retains 87% of its peak EQE at 100 mW mW sr⁻¹ m^-2, whereas the control drops to 46% at 25 mW mW sr⁻¹ m^-2, demonstrating markedly suppressed efficiency roll-off. g Capacitance–voltage (C–V) characteristics, indicating reduced charge accumulation and more uniform transport in the ordered film. h Schematic of the degradation mechanism in the control device: the disordered QD layer promotes charge accumulation, localized heating and leakage at the electron transport layer (ETL) and hole transport layer (HTL) interfaces. i Schematic of the target device: the ordered QD assembly enables uniform charge distribution, minimized heat generation and suppressed leakage, improving operational stability.

The resulting LEDs exhibit substantially suppressed leakage current compared to control devices, as shown in the current density-voltage characteristics (Fig. 5c). At identical voltages, target devices delivered significantly enhanced radiance: the maximum radiance increased nearly ten-fold from 33.86 mW sr⁻¹ m^-2 for control devices (at 2.5 V) to 323.31 mW sr⁻¹ m^-2 for target devices (at 2.2 V). Correspondingly, target devices achieved a peak EQE of 3.75%, approximately ten-fold higher than control devices (0.38%) (Fig. 5d). Moreover, target LEDs exhibited markedly reduced efficiency roll-off across the measured voltage range (Fig. 5d). These performance metrics represent state-of-the-art among erbium-doped electroluminescent devices. The combination of 3.75% EQE, 2.2 V turn-on voltage, and 323.31 mW sr⁻¹ m^-2 maximum radiance establishes a new benchmark for solution-processed optical communication-band emitters, significantly surpassing previously reported systems. To evaluate device reproducibility, 20 independently fabricated devices were measured, yielding an average EQE of 3.70% (Supplementary Fig. 19).

We attribute these performance improvements to hierarchical structural control at multiple length scales. At the nanocrystal level, target QDs exhibit reduced defect densities, as evidenced by longer photoluminescence lifetimes (31.3 ns vs. 15.4 ns, Supplementary Fig. 20). To directly probe the out-of-plane transport of the QD films, we performed vertical SCLC measurements using electron-only and hole-only devices (Supplementary Fig. 21). The target film exhibits substantially lower trap-filled limit voltages than the control in both carrier channels, corresponding to reduced trap densities from 3.29$\times$10¹⁷ to 1.5$\times$10¹⁷cm^-3 for electron transport and from 3.03$\times$10¹⁷ to 1.25$\times$10¹⁷cm^-3 for hole transport. The carrier mobility is also improved in the target film⁴⁵. These electrical transport improvements are further corroborated by TA measurements comparing the excited-state dynamics of disordered and ordered CsPbCl₃:Er³⁺ films (Supplementary Fig. 22). The ordered film shows slower bleach recovery, with τ_avg prolonged more than threefold from 55.34 ps (disordered) to 175.50 ps (ordered). In addition, target films form close-packed, mesoscale ordered assemblies with uniform emission (Supplementary Figs. 23–25). This structural ordering significantly suppresses leakage current and enhances charge injection efficiency, as reflected in the substantially improved current density–voltage characteristics (Fig. 5c).

Benefiting from these improvements, we measured operational stability by applying a constant current density of 10 mA cm^-2 and monitoring normalized radiance over time. Target devices exhibited an operational half-lifetime (T₅₀) of 197 min, approximately seven times longer than control devices (28 min) (Fig. 5e). The superior stability of target devices is further demonstrated across varying radiance levels: at radiances from 10 to 100 mW sr⁻¹ m^-2, target devices consistently maintained normalized EQE values above 87%, while control devices showed significantly lower stability, dropping to 46% at 25 mW sr⁻¹ m^-2 (Fig. 5f).

To understand the underlying mechanisms, we performed capacitance–voltage (C–V) measurements to gain deeper insights into these stability improvements (Fig. 5g). Target devices display suppressed charge accumulation in the built-in potential region and delayed trap filling compared to control devices, which show severe charge accumulation, indicating reduced defect densities and more uniform carrier distribution in target devices^46,47. We further examined the degradation mechanisms using schematic diagrams (Fig. 5h, i): in control devices, the disordered quantum dot layer promotes charge accumulation, localized heating, and leakage pathways, leading to rapid performance degradation. In contrast, target devices with ordered quantum dot assemblies exhibit uniform charge distribution, minimized heat generation, and suppressed leakage, resulting in enhanced operational stability. Our hierarchical ordering strategy establishes a chemistry-to-device pathway to transform Er³⁺-doped QDs from inefficient emitters into viable electrical pumping sources at optical communication wavelengths, achieving simultaneous improvements in efficiency, radiance, operating voltage, and device lifetime.

Discussion

In summary, we implement a synergistic slow-release halide chemistry coupled with in-situ amide-ligand–directed assembly, which stabilizes monomer supply to achieve size focusing and low defect density, and—via N–H···O = C hydrogen-bond networks—induces oriented assembly of QDs to construct mesoscale ordered Er³⁺-doped CsPbCl₃QD films. The device improvements originate from the coupled optimization of transport and recombination: the ordered assembly lowers trap density, suppresses charge accumulation, and prolongs the excited-state lifetime available for Er³⁺ sensitization, collectively delivering higher EQE, radiance, and operational stability. Consequently, the devices deliver efficient and load-tolerant electroluminescence at 1.55 µm, with a peak EQE of 3.75%, 87% EQE retention at 100 mW sr⁻¹ m^-2, T₅₀ of 197 min, and a maximum radiance of 323.31 mW sr⁻¹ m^-2 (versus 33.86 mW sr⁻¹ m^-2 for controls; ~10 times increase). This work identifies the key control parameters for lanthanide near-infrared QDs and provides transferable design rules, suggesting that the chemistry-to-order paradigm can be extended to other lanthanides/halides and heterostructures, advancing narrowband, low-voltage, long-lifetime electrically driven NIR emitters across the telecom band, with potential utility in optical sensing and integrated near-infrared photonics.

Methods

Materials

Cesium acetate (CsOAc, 99%, Adamas), Lead (II) acetate trihydrate (Pb(OAc)₂·3H₂O, 99%, Sigma-Aldrich), Erbium(III) triacetate hydrate (Er(OAc)₃·4H₂O, 99.99%, Aladdin), Myristoyl Chloride (Myr-Cl, 98%, TCI), Trimethylchlorosilane (TMS-Cl, 99%, Sigma-Aldrich), Oleylamine (OAm, technical grade 70%, Sigma-Aldrich), Oleic acid (OA, technical grade 90%, Sigma-Aldrich), 1-Octadecene (ODE, technical grade 90%, Sigma-Aldrich), PEDOT:PSS solution (Clevios PVP AI 4083), poly-TPD, and PO-T2T were obtained from Xi’an Polymer Light Technology Corp. All materials were utilized without further purification.

Synthesis of undoped and Er-doped CsPbCl₃ quantum dots

Quantum dots were synthesized via a modified hot-injection protocol. The reaction used 18-octadecene (ODE, 5 mL) as solvent, with oleic acid (OA) and oleylamine (OAm) added at a volume ratio of 2:1. The halide precursor was prepared by adding equal volumes of trimethylsilyl chloride (TMS-Cl) and myristoyl chloride (Myr-Cl) into 2 mL ODE and mixing thoroughly. For the metal precursor, cesium acetate (0.28 mmol) was used as the A-site source. For the undoped route, lead acetate was used as the sole B-site precursor; for the Er-doped route, lead acetate and erbium acetate were introduced at a 1:1 molar ratio. The reaction mixture was degassed, purged with nitrogen three times, and then heated under vacuum to 120 °C for 60 min until the solution became clear. Vacuum was released and, under an inert atmosphere, the temperature was raised to 250 °C. Upon reaching a uniform temperature, 0.9 mL of the halide precursor solution was swiftly injected; after 10 s of growth, the flask was quenched in an ice–water bath to terminate the reaction, affording uniform nanocrystals.

Post-processing of QDs

Add an antisolvent to the crude quantum dot solution. Centrifuge the mixture to collect the precipitate, then dissolve it in toluene. Repeat the above steps to obtain a purified CsPbCl₃:Er³⁺ solution. Store the quantum dot solution at 4 °C for subsequent use.

Device fabrication

To fabricate the QLEDs, ITO-coated glass substrates were sequentially cleaned in ultrasonic baths of deionized water, acetone, and 2-propanol for 60 min in each solvent. Upon drying under a nitrogen stream, the substrates were subjected to UV-ozone treatment for 15 min. Subsequently, a PEDOT:PSS layer was deposited via spin-coating at 4000 rpm for 30 s, and then baked on a hotplate at 150 °C for 15 min. After the substrate was cooled to room temperature, a solution of 8mg/mL poly-TPD chlorobenzene was spin-coated onto the PEDOT: PSS layer at 4000rpm for 40s, baked for 15min at 100°C under an inert nitrogen atmosphere. Then, the QD films are coated using the method above. Finally, ∼60 nm of POT2T, ∼2 nm of Liq, and ∼80 nm of Al was deposited using a thermal evaporation system (Suzhou Fangsheng FS380-S12) through shadow masks under a high vacuum (<10^-4 Pa).

Optical characterization

Ultraviolet–visible (UV–Vis) absorption spectra were acquired using a PerkinElmer Lambda 950 spectrophotometer. For photoluminescence (PL) measurements, spectra were recorded on an NIR-VIS fluorescence spectrometer (FL3). PL decay profiles were determined via a PL-TCSPC fluorescence lifetime system (Hamamatsu Photonics Co., C12132-38), while additional transient PL decay data were gathered using a HORIBA-FM-2015 instrument.

Femtosecond transient absorption (TA): TA measurements were performed using a Helios ultrafast pump–probe system coupled with a Coherent regenerative amplified laser. The excitation source was an Astrella Ti: sapphire amplifier (Coherent), which emitted 100 fs pulses at 800 nm with a repetition rate of 1 kHz. The fundamental output was split into two distinct optical paths via a beam splitter. One fraction of the beam was routed into an optical parametric amplifier (TOPAS, Coherent) to produce 320 nm pulses used for pump excitation. The remaining portion was focused onto sapphire and CaF₂ plates to produce a white-light supercontinuum probe spanning 300–650 nm. The pump pulses were modulated at 500 Hz using a mechanical chopper.

Structural characterization

X-ray diffraction (XRD) patterns were recorded using a Bruker AXS D8 diffractometer equipped with a Cu Kα radiation source (λ = 1.54178 Å). High-resolution transmission electron microscopy (HRTEM) images were acquired using an FEI Talos F200X microscope operated at an accelerating voltage of 200kV, followed by image analysis with the DigitalMicrograph software package. Atomic-resolution energy-dispersive spectroscopy imaging was acquired on a Thermo Fisher Spectra 300 scanning transmission electron microscope in STEM mode, with operating parameters of 120 kV acceleration voltage and 50 pA screen current. To reduce electron-beam-induced damage to the perovskite QDs, all alignments and focusing were performed away from the areas for imaging.

Film-surface characterization

The surface morphology was characterized using a Hitachi SU8230 scanning electron microscope (SEM) operated at an accelerating voltage of 10 kV. Atomic force microscopy (AFM) images were acquired via a Cypher-S system (Asylum Research, Oxford Instruments, UK). For electronic structure analysis, ultraviolet photoelectron spectroscopy (UPS) and X-ray photoelectron spectroscopy (XPS) were carried out on a KRATOS AXIS Ultra DLD spectrometer maintained at a base pressure of 10^-9torr. Grazing incidence wide-angle and small-angle X-ray scattering (GIWAXS/GISAXS) experiments were performed using a Xeuss 3.0 instrument at the Vacuum Interconnected Nanotech Workstation (Nano-X), Suzhou Institute of Nano-Tech and Nano-Bionics (Chinese Academy of Sciences), with the incident angle fixed at 0.2°.

Density functional theory calculations

First-principles computations were carried out utilizing the Vienna Ab-initio Simulation Package (VASP). To describe electron-ion interactions, we adopted the Projector Augmented Wave (PAW) method. The exchange-correlation potential was treated within the Generalized Gradient Approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) functional, with DFT-D3 corrections applied to account for van der Waals forces. Brillouin zone sampling employed a 4 × 4 × 4 k -point mesh. Structural relaxations were performed with a plane-wave cutoff energy of 600 eV, continuing until the energy difference dropped below 10-5 eV and the force on each atom was less than 0.02 eV/Å.

To determine the relative stability of different configurations of the Er occupying interstitial or subsituational sites, we calculated the Er formation energies (Eform) in the following three cases:

1. CsPbCl₃ substituting Er atom for Pb atom on the Pb site (S_Er-CsPbCl₃) E_form = E(S_Er-CsPbCl₃) + μ_Pb − E(CsPbCl₃) – μ_Er

2. CsPbCl₃ substituting Er atom for Cs atom on the Cs site (T_Er-CsPbCl₃) E_form = E(S_Er-CsPbCl₃) + μ_Cs − E(CsPbCl₃) – μ_Er

3. CsPbCl₃ with Er atom entering interstitial site of (I_Er-CsPbCl₃) E_form = E(I_Er-CsPbCl₃) − E(CsPbCl₃) – μ_Er

Characterization of NIR QLEDs

The current density–voltage (J–V) profiles were acquired using a computer-integrated Keithley 2400 source meter. To characterize the electroluminescence (EL) spectra and external quantum efficiency (EQE), we employed a calibrated Hamamatsu PMA−12 photonic multichannel analyzer system (models C10027-01 for 360–950 nm and C10028-01 for 950–1600 nm). This setup was coupled with a 3.3-inch integrating sphere to collect forward emission, driven by a regulated power supply. Absolute radiance calibration was performed using a Photo Research PR-745 spectroradiometer (380–1060 nm), assuming Lambertian emission. The visible-range EL spectrum was recorded using a PR745 instrument, while the NIR EL spectra were collected using an Ideaoptics NIR17S spectrometer. The EQE was determined using a calibrated InGaAs photodetector sensitive in the 800–1700 nm range; the visible host emission at ~410 nm is therefore excluded from the EQE calculation. The normalized EQE retention shown in Fig. 5f was extracted from the standard forward J–V–L sweep and represents EQE roll-off as a function of radiance.

Author contributions

H.H.L. and J.L.P. contributed equally to this work. L.S.L. and Y.K.W. conceived and directed the project. H.H.L. synthesized the Er³⁺-doped CsPbCl₃ quantum dots and performed characterizations under the supervision of L.S.L. and Y.K.W. H.H.L. and J.L.P. fabricated the QLEDs under the supervision of L.S.L and Y.K.W. Y.Y.P. performed the device characterizations under the guidance of Y.K.W. F.Z. conducted transmission electron microscopy (TEM) imaging of quantum dots and assisted with device testing. H.B.R. and W.S.S. assisted with device characterizations. The funding was obtained by L.S.L. and Y.K.W. H.H.L. wrote the manuscript under the supervision of L.S.L. and Y.K.W. All authors discussed the results and commented on this manuscript.

Peer review

Peer review information

Nature Communications thanks Zhi-Kuang Tan, Tae-Woo Lee, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

Y.K.W. discloses support for the research of this work from the National Natural Science Foundation of China (NSFC) [grant numbers 62522514 and 62474119], Gusu Innovation and Entrepreneurship Leading Talent Program [grant number ZXL2024367], and National Key Research and Development Program of China [grant number 2024YFA1209500]. L.S.L. discloses support for publication of this work from the National Natural Science Foundation of China (NSFC) [grant number 62175171], Science and Technology Development Fund, Macao SAR [grant number 0072/2024/RIB2]. W.S.S. discloses support for publication of this work from the China Postdoctoral Science Foundation [grant number 2025M780125], Basic Research Program of Jiangsu [grant number BK20250810], and Jiangsu Funding Program for Excellent Postdoctoral Talent [grant number 2025ZB341]. H.H.L., J.L.P., Y.Y.P., F.Z., and H.B.R. declare no relevant funding.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary files. Any additional information related to the study is available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75429-3.