Quantum dot light-emitting diodes (QD-LEDs) have emerged as a transformative technology in the fields of displays, lighting, and lasers due to their exceptional color purity, tunable emission spectra, and high luminescence efficiency^[1−12]. The unique electronic and optical properties of quantum dots (QDs) are dictated by the quantum confinement effect, wherein reducing the particle size below the exciton Bohr radius leads to significant modifications in energy levels and carrier dynamics^[13−15]. This tunability enables precise control over emission wavelength and efficiency, making QDs ideal for diverse optoelectronic applications^[16−19]. Notably, quantum dots have recently made their mark in the commercial display sector, appearing as integral films within liquid crystal display (LCD) panels. Meanwhile, QLEDs that utilize QDs as the emissive layer have achieved brightness and external quantum efficiency (EQE) levels comparable to those of OLEDs. However, the quantum confinement effect also introduces size-dependent challenges that critically influence charge transport, exciton dynamics, and overall device performance^[20−22]. The interplay between the exciton Bohr radius (r_B) and the size (or volume) of QDs fundamentally governs the strength of quantum confinement effects. When considering the scenarios of highly confined quantum dots, weakly confined quantum dots, and nanocrystals (NCs) without confinement in the same size distribution range, Fig. 1(a) shows that the energy level distribution near the band edge (ΔE) depends on the degree of quantum confinement, which determines how energy levels are discretized. QDs with broad size distributions (Δd) exhibit a spread of HOMO (highest occupied molecular orbital) and LUMO (lowest unoccupied molecular orbital) energy levels due to their size variations, resulting in inhomogeneous broadening of absorption and emission spectra, whereas changes in NC size have little impact on energy levels due to the absence of confinement effects (Fig. 1(b)).

Cesium lead halide (CsPbX₃, X = Cl, Br, I) QDs represent a new frontier in QD-LED research, offering advantages such as high photoluminescence quantum yields (PLQY), narrow emission linewidths, and defect tolerance^[23−26]. Their size-dependent properties provide a unique opportunity to optimize both optical and electronic characteristics for LEDs^[27−31]. Nevertheless, the practical implementation of CsPbI₃ QDs in LEDs faces significant hurdles. CsPbI₃, being an ionic crystal with relatively low formation energy, exhibits high sensitivity to environmental factors such as humidity, temperature fluctuations, and light exposure. These conditions significantly influence its long-term stability and reliability. Therefore, choosing CsPbI₃ as the research subject not only aids in addressing the challenges associated with its poor device stability but also holds significant implications for advancing its commercialization process. This study aims to bridge the gap between fundamental quantum confinement physics and the practical design of QD-LEDs by systematically investigating the impact of QD size on charge transport, exciton recombination dynamics, and device efficiency.

Using CsPbI₃ QDs as a model system, we provide a comprehensive analysis of size-dependent trade-offs, discuss advanced strategies for overcoming performance bottlenecks, and propose new design rules for next-generation QD-LEDs. Smaller QDs, though exhibiting high initial efficiencies, suffer from increased non-radiative losses and efficiency roll-off at higher carrier densities due to Auger recombination and surface defects. Larger NCs, on the other hand, demonstrate improved charge transport and higher brightness under high current densities but require innovative strategies to achieve similar levels of efficiency. By integrating insights from optical physics, materials science, and device engineering, this work lays the foundation for scalable and efficient QD-LED technologies.

As CsPbI₃ has a r_B of 6 nm, we selected particles with a diameter of 12 nm as the weakly confined quantum dots, which we named QD in short. As another research object, NCs without confinement were chosen, and were given the name NC. We obtained CsPbI₃ colloidal particles by using the hot-injection method (see Methods), and QD and NC were obtained following size separation, which was achieved by centrifugation. Due to the fact that QD and NC are formed from the same batch of synthesis, their optical properties are similar. The colloidal CsPbI₃ has a high photoluminescence (PL) quantum yield over 90%.

Fig. 2(a) and 2(b) present typical transmission electron microscopy (TEM) images of QD and NC, respectively, revealing rather monodisperse cubic-shaped particles. The high-resolution TEM images with typical planes (100) are well able to display their crystal structures due to their high degree of crystallinity, as shown in Fig. 2(c) and 2(e)^[32]. According to Fig. 2(d) and 2(f), the edge lengths of QD and NC are 11.4 ± 2.0 and 15.8 ± 1.9 nm, respectively. This indicates they have the same size distribution range, providing a basis for further discussion. The X-ray diffraction (XRD) analysis confirms that both QD and NC possess a phase pure perovskite structure. The diffraction peaks observed at 14.6°, 20.65°, and 29.1° correspond to the (100), (110), and (200) crystal planes, respectively (Fig. S1)^{[33, 34]}.

The UV−vis absorption and PL spectra of CsPbI₃ particles dissolved in octane are displayed in Fig. 3(a). Fig. S2 shows a photograph of a high purity colloidal solution showing bright red emission (365 nm excitation wavelength). The 4.4 nm increase in diameter resulted in a 3 nm PL redshift from 687 nm for QD to 690 nm for NC. QD and NC have full width at half maximum (FWHM) values of 35 and 33 nm, respectively, both of which are quite narrow. Due to the greater sensitivity of energy level distribution to size changes, the FWHM of QD has increased slightly. Weakening quantum confinement causes a longer exciton lifetime, as well as increased particle volume decreasing the chances of particles from getting charged, both contributing to a longer PL lifetime (Fig. 3(b))^[29]. In Table S1, the fitting details are shown and it is revealed that the PL decay is slowed down from 110 to 79 ns. These findings underscore the confinement-dependent tunability of optical and charge dynamics in CsPbI₃ systems.

By analyzing ultraviolet photoelectron spectroscopy (UPS) results (Fig. 4(a) and 4(b)) and Tauc plots (Fig. 4(c)) for these CsPbI₃ films, we confirmed the energy level structures of these films prior to examining their charge transport behavior. According to the cutoff region, NC and QD have Fermi energy levels (E_F) of −3.7 and −3.2 eV, respectively. The band tail states in QD are quite clear (marked in yellow) while in NC they are negligible, indicating a broader energy level distribution around the E_F of QD, in accordance with confinement-dependent optical properties and charge dynamics. According to the valence region, the difference between HOMO and E_F is 1.3 eV for NC and 1.8 eV for QD. The Tauc plots reveal bandgaps of 1.78 and 1.80 eV for NC and QD, respectively. Thus, we can derive the HOMO values for NC and QD to be −5.2 eV for both and the LUMO values for NC and QD to be −3.22 and −3.2 eV, respectively. Having acquired the above information, we can now map energy level structures and clarify how to design devices to investigate charge transport.

Both QD and NC have a HOMO of −5.2 eV, so when we investigate their hole transport, the confinement-dependent energy level distribution will be the only experimental variable, which will provide a more convincing proof than the one obtained by investigating electron transport. The hole transport was then studied using a designed device structure of ITO/poly(ethylenedioxythiophene):polystyrenesulfonate (PEDOT:PSS, 40 nm)/CsPbI₃ (80 nm)/MoO_x/Au (Fig. 4(d)), whose energy band diagram can be seen in Fig. 4(e). The hole-only device allows Ohmic contact between electrodes and functional layers, thereby facilitating the efficient transport of holes within CsPbI₃ films and providing an almost essential description.

The current density−voltage (J−V) curves of hole-only devices reveal how energy level distribution affects hole transport (Fig. 4(f)). NC-based devices have significantly higher current density than QD-based devices. As a result of greater energy misalignments between adjacent particles, there is an energetic barrier for charge hopping in QD than in NC, which reduces the efficiency of charge transport because carriers must overcome these barriers by thermally assisted hopping or tunneling. Another reason is that the wider size distribution in QD causes significant HOMO variation, leading to trap states or energy barriers, which cause hole losses and reduce film conductivity. Furthermore, for QDs with quantum confinement, a larger quantum level spacing introduces the Coulomb blockade effect^{[35, 36]}, which requires additional energy to add a charge that further restricts hole transport. QD-based devices exhibit a higher trap-filled limit voltage than NC-based devices, increasing from 1.07 to 1.25 V, further demonstrating that the size distribution introduces more trap states into QD than NC. The mixture of NC and QD, designated MIX, exhibits hole transport capabilities between them, indicating that NC does not introduce additional energy levels to QD, indicating that NC's LUMO deviation is within QD's LUMO deviation range. Quantum confinement effect transitions from being an important factor in confined QD to being negligible in bulk-like NC, causing difficulties in charge transport.

The deviation of HOMO and LUMO levels in particles has a profound influence on charge transport in films, which ultimately impacts the performance of LEDs. Efficient charge transport is crucial for the operation of LEDs, as it ensures that electrons and holes can recombine effectively to emit light. Poor charge transport can lead to energy losses, reduced brightness, and shorter device lifetimes^{[37, 38]}. The LED structure was designed based on the energy level structure of CsPbI₃ as follows: ITO/PEDOT:PSS/poly(N,N′-bis(4-butylphenyl)-N,N′-bis-(phenyl)benzidine) (poly-TPD)/CsPbI₃/2,4,6-tris[3-(diphenylphosphinyl)phenyl]-1,3,5-triazine (PO-T2T)/LiF/Al (Fig. 5(a)). Fig. 5(b) shows a schematic of the flat-band energy level diagram of our LEDs. A precious optimization can establish Ohmic contact between electrodes and charge transport layers. A 0.2 eV barrier is present for hole injection from poly-TPD into CsPbI₃, which is almost the same as the 0.3 eV barrier for electron injection from PO-T2T into CsPbI₃, which helps ensure charge balance within the emitters.

The energy alignment between emitter HOMO-LUMO levels and the injection layers is crucial for efficient charge injection. Quantum confinement dictates the energy levels of QD, in which charge carriers must encounter greater injection barriers when particles are smaller in size. It is evident from the LED J−V curves (Fig. 6(a)), that the current density increases much more rapidly for NC LEDs than for QD LEDs and MIX LEDs. NC LED is characterized by a low turn-on voltage of 2 V because of the efficient charge injection process, the lowest among the three types of LEDs. NC have weaker quantum confinement, resulting in a narrower bandgap compared to smaller QD, which reduces the energy required for charge carriers to overcome the bandgap and recombine radiatively, lowering the turn-on voltage. According to Fig. 6(b), the brightness of NC LEDs increases more rapidly than other LEDs, further indicating that NC LEDs have the smoothest charge transport and injection. The peak brightness of NC LED, MIX LED, and QD LED are respectively 1857, 755, and 377 cd·m⁻². NC LED is able to produce higher peak brightness due to the suppressed charging phenomenon and decreased thermal generation that result from this.

Fig. 6(c) illustrates that QD LEDs exhibit a high external quantum efficiency (EQE) of 25% at a current density of 0.1 mA·cm⁻², but the EQE quickly decreases to 10% at a current density of 30 mA·cm⁻². As a comparison, NC LEDs exhibit EQE curves that keep growing along with the current density, reaching their maximum level of 25% at 4.3 mA·cm⁻² and then gradually declining to 10% until 222 mA·cm⁻². As quantum confinement enhances radiant recombination, it also increases particles' likelihood of becoming charged, particularly in smaller particles, which enhances the emission under low current densities, but limits the performance of the device under high current densities.

The Mott−Schottky plots of capacitance−voltage characterization of LEDs in Fig. 6(d) illustrate the built-in voltages for LEDs, which decrease from 3.7 to 2.6 V for QD LED to NC LED, in accordance with the improved charge injection and transport as well as the reduced turn-on voltage, which further illustrates the physical mechanism of device performance changes. Note the Mott−Schottky plots are obtained in air while LED performances are obtained in a N₂-filled glove-box, indicating the evolving trend of the built-in voltage is significant, but its absolute value is meaningless.

Fig. 7(a) shows LED electroluminescence (EL) spectra. The inset depicts a photograph of the LED devices operating at their maximum brightness. Notably, NC-based devices exhibit superior brightness compared to those QD-based LEDs. In NC LEDs, the EL peaked at 688 nm with a constant FWHM of 33 nm. QD LED has an EL peaking at 684 nm, with corresponding spectra broadening on the short wavelength side, with FWHM extending from 34 to 36 nm along with charge density increases. It reveals the filling processes of relatively lower energy states within the wider distribution of energy states around LUMO. MIX LED also has EL peaking at 686 nm with FWHM values increasing from 34 to 35 nm. The larger NC has demonstrated a lower probability of being charged than the QD, which results in less thermal generation during operation, thereby extending the device's lifespan. As shown in Fig. 7(b), at an initial brightness of 100 cd·m⁻², the device’s half-lifetime increased from 3.5 min to 18.3 h. We evaluated the thermal stability of the NC and QD by heating both films to 120 °C for a period of 5 min and monitoring the changes in their PL intensities before and after the temperature increased (Fig. S3). Our results indicated that the PL intensity of both films diminished upon exposure to 120 °C, specifically, the PL intensity of the NC film reduced to 22% of its initial value, while that of the QD film decreased to 9% of its original intensity. Although there was a notable difference in PL retention between the two films, this disparity does not fully account for the more pronounced gap observed in the stability of corresponding devices. Consequently, we conclude that the differences in thermal stability contribute only marginally to the variations in device performance.

The quantum confinement effect serves as a double-edged sword in QD-based LEDs. By using QD and NC whose particle size is comparable to or larger than the exciton Bohr diameter of CsPbI₃, we achieved a significant modification of energy levels due to the spatial restriction of electron and hole wavefunctions. With a similar size distribution, QD demonstrates a more significant broadening of HOMO and LUMO energy levels and spectral features than NC. QD delivers superior emission efficiency at lower current densities but faces limitations at high excitation, while NC excels at achieving high brightness at elevated current densities. We believe that utilizing quantum dots of appropriate size, balancing carrier injection, and enhancing the efficiency of light output extraction may be conducive to achieving QLED with high efficiency, high brightness, and long operational lifespan. This study demonstrates how quantum confinement effects enable tunable optical properties and enhanced radiative recombination, while introducing challenges in charge transport and non-radiative losses. To realize the full potential of QD-LEDs, it is critical to adopt targeted design rules that address the unique challenges posed by the size-dependent properties of QDs. These principles should aim to balance charge injection, mitigate non-radiative losses, and enhance stability while leveraging the tunable properties of QDs.