Effect of Tellurium Doping on Optoelectronic Properties of Blue ZnTeSe Quantum Dots

Quantum dot light emitting diodes (QD-LEDs) have high potential to be used in next-generation displays. The main challenge in the commercialisation of QD-LEDs is the improvement of the eciency and stability of blue QD-LEDs. The most promising emitter for blue QD-LEDs is ZnTeSe QD, however it is concerned that tellurium causes spectral broadening and eciency drop. In this work, we demonstrate the effect of Te clusters on the electronic structures of blue ZnTeSe QDs through rst principle calculation. The photo dynamics at the ensemble and single dot level show that the lifting of hole-state degeneracy in Te-doped structures causes slow relaxation and peak emission shift. Furthermore, the strong connement of hole wave functions in Te clusters improves the eciency and stability of QD-LEDs by suppressing QD luminescence quenching under electrical bias. The understanding of the correlation between the photophysical nature of Te-doped QDs and device performance can provide a basis for designing blue-emitting QD structures that are suitable for practical QD-LEDs. The of examined through DFT calculations. The photo dynamics at the ensemble and single dot levels revealed that Te clusters formed localised hole states. The Te clusters in ZnTeSe-based QDs improved the eciency of the QD-LEDs by suppressing QD luminescence quenching under electrical bias because of strongly bound localised hole states. Furthermore, quantitative analysis showed that localised Te-hole states affected the EL spectral features of the QD-LEDs under multi-exciton generation conditions. Our results provide a fundamental understanding of ZnTeSe-based blue QDs for practical application to QD-LED displays.


Introduction
The development of e cient and stable blue-emitting uorophores is one of the challenges for nextgeneration quantum dot light emitting diode (QD-LED) displays. Although QD-LEDs have numerous advantages over conventional technologies 1,2 , the performance of blue QD-LED devices is worse than that of red and green QD-LED devices [3][4][5] . This is partly due to the di culties in synthesising high quality blue quantum dot (QD) materials, which intrinsically have more surface defects compared to red or green QD materials owing to the small size of CdSe or InP 4,6 . ZnSe QDs have a wide energy band gap. Thus, it is necessary to add a small amount of tellurium (Te) to tune the emission at a blue wavelength of 460 nm. Numerous works have investigated ZnTeSe QDs, and highly e cient blue QDs with near-unity photoluminescence (PL) quantum yield (QY) have been recently reported 7 . However, still there are doubts on using Te for the blue QDs based on the expectations that it is more vulnerable to oxygen than selenium (Se) and induces multiple trap states that increase the spectral width and decrease the QY 8 .
Thus, to design blue QD structures that are suitable for LED applications, it is necessary to understand the role of Te doping in ZnTeSe with respect to its photonic properties and correlation with device performance. There is no strong relationship between the PL QY and external quantum e ciency (EQE) of state-of-the-art QD-LEDs because the current EQE strongly depends on the electron/hole (e/h) charge balance rather than the luminescence of QDs 9,10 . In addition, previous studies have reported that it is critical to alleviate the accumulation of charges inside QDs by preparing a thick shell or alloy shell for achieving a high PL QY, particularly under multi-exciton generation conditions 11,12 . Therefore, to develop highly e cient and stable blue QD-LEDs, we believe that it is important to address the issues of e/h behaviours such as e/h overlap under bias, exciton recombination inside ZnSe-based QDs, and their complex interactions with adjacent conducting layers in LED devices. In this work, we examine the role of Te doping in blue-emitting ZnTeSe QDs to create highly e cient and stable electroluminescent (EL) devices and investigate the origin of the improvements in the device performances. The electronic structures of Te-doped ZnTeSe QDs signi cantly in uence PL and EL properties by forming localised hole states close to the valence band maximum (VBM); this eventually enhances the quantum con nement effect. The spatial clustering effect of Te in ZnTeSe QDs is systematically shown via density functional theory (DFT) calculations. Additionally, hole relaxation properties of blue emitting ZnTeSe QDs are experimentally con rmed by measuring transient absorption (TA) dynamics and photo dynamics at the ensemble and single dot levels for the rst time. It is newly suggested that Te-induced localised hole states facilitate single exciton recombination by suppressing PL quenching under a strong electric eld.
Furthermore, the EL spectral shift of the blue ZnTeSe based QD-LEDs is compared with the changes in the PL of the QDs with different Te concentrations at similar exciton densities.

Zntese Qds With Localised Te States
ZnTeSe cores with Te concentrations of 0%, 3%, 5%, and 7% and respective ZnTeSe/ZnSe/ZnS core/multi-shell QDs were prepared according to previously introduced methods 13,14 . Controlled amounts of Te and a Se precursor were simultaneously added during the preparation of a core. Then, the ZnSe/ZnS shell was grown on the core in the presence of halide additives. Figure 1a shows the schematic and high-resolution transmission electron microscopy (HR-TEM) image of a ZnTeSe/ZnSe/ZnS QD and the diffraction pattern observed in the [100] direction. All the QDs showed pure zinc blende crystalline structures (Supplementary Figure 1), and the core/shell QDs exhibited high PL QYs of 93-97% with average sizes of ~10 nm (Supplementary Figure 2). The detailed optical properties, elemental compositions, and structural information of the QDs are presented in Supplementary Table 1. As previously reported 15,16 , the most noticeable effect of Te doping on the optical properties of the ZnTeSe/ZnSe/ZnS QDs was the red shift in the PL emission from 440 nm to 461 nm and an increase in the spectral width from 12 nm to 36 nm as the Te concentration increased to 7% (Supplementary Figure   3). This was because of the narrowing of the energy band gap due to the contribution of the ZnTe energy level. However, the concurrent changes in the Stokes shifts (from 2 nm to 8 nm) and average transient PL lifetimes (from 18 ns to 55 ns) with the Te concentration (Supplementary Table 1) implied that the energy transitions in the electronic structures of the ZnTeSe QDs were fundamentally altered from the simple combination of the ZnSe and ZnTe electronic structures. In CdSe doped with a few Te substitutional impurities, the Stokes shift, radiative decay time, and biexciton binding energy depended on the QD size owing to the carrier localisation around the Te dopant 2 . We used the density functional method and the projector-augmented wave pseudopotential provided by the Vienna ab initio simulation package to understand the electronic structure and carrier localisation effect of Te doping 17 . The size of atomic clusters and the number of Te dopants in ZnTeSe QDs were similar to experimental values. Speci cally, for the ZnSe core, an atomic cluster model with a diameter of 3.6 nm was constructed with 532 Zn atoms and 555 Se atoms, and we randomly placed Te atoms at Se positions in the Te-doped structures. Figure  1b shows the representative atomic models of the ZnSe and ZnTeSe core QDs with 7% Te and the corresponding energy levels close to the band edges. Te doping induced an upshift of the VBM while maintaining conduction band minimum (CBM). Furthermore, the clustering size (number of Te atoms in a cluster) affected the degree of changes in the VBM. We calculated the energy band gap of ZnTeSe QDs at different Te concentrations and clustering sizes and compared it with experimental values (Figure 1c).
For each given Te concentration, band gap calculations were performed for selected Te distributions with high probability among randomly generated 30,000 atomic con gurations. The maximum number of Te atoms per cluster determined energy band gap, which accorded to the experimental values changed from 3.36 eV to 3.21 eV. The locations and shapes of Te clusters also affected band gap to result in the discrepancies of calculated values for same sized cluster (Supplementary Data 1). The calculated oscillator strength and absorbance spectra also showed the same trend of band gap decrease, and the degeneracy close to the lowest states was lifted (Supplementary Figure 4). Moreover, the discrete energy states close to the VBM developed more as the clustering size increased with the Te concentration.
Therefore, the localised Te energy level of ZnTeSe QDs can play an important role in determining photophysical and EL properties.
The effect of the Te cluster on the energy transition was veri ed by measuring the fast temporal evolutions of TA dynamics as a function of the Te concentration in ZnSeTe/ZnSe/ZnS core/shell QDs ( Figure 1d). To ensure the measurement of relaxation without accumulated charges, pump uence was controlled at a low excitation level (50 µW), which corresponded to an average exciton occupancy ⟨N x ⟩ of 0.5. Consecutive measurements were performed with (A) and without (A 0 ) a pump pulse to obtain the transient absorbance difference, ΔA = A -A 0 , which represented the population of excited carriers at a given energy in the delay time between the pump and probe pulses. The photo generated hot carriers in QDs relaxed to the band-edge state within a few picoseconds, which was accompanied by a rise and decay in TA spectra. Figure 1e shows the temporal evolutions of ΔA for the ZnSeTe/ZnSe/ZnS core/shell QDs with 0%, 3%, 5%, and 7% Te. In ZnSe QDs, a distinct band structure was observed in a short period (t We further investigated the optical properties of individual QDs to exclude the effect of distributions. Figure 2a shows the PL spectra of a representative single dot (average and 1 s frame) and the ensemble for the ZnTeSe/ZnSe/ZnS QDs with different Te concentrations. For all samples, the shapes of the accumulated spectra of more than 30 single dots were almost identical to that of the respective ensemble (Supplementary Figure 6). It is worth noting that the spectral width of the ensemble and single dot PL spectra increased with the Te concentration ( Figure 2b). This implied that the major reason for the spectral broadening was Te-induced hole states, regardless of heterogeneous particle distribution or spectral diffusion. Furthermore, the PL lifetime of the ensemble and single dot decreased with the peak energy owing to the localised Te states ( Figure 2c). As these characteristics of single dot analysis were the same as those observed in ensembles, it was clear that localised Te states were the origin of the spectral modulations instead of the heterogeneity of particles. In addition, the surface defect emission for Te-doped ZnTeSe core QDs (Supplementary Figure 3) was considerably suppressed at low temperatures, which suggested strong spatial con nement in the core owing to localised Te states.

Effect Of Localised Te States On Qd-leds
The effect of the Te-induced energy levels on the QD-LEDs was investigated by fabricating unit devices using the ZnTeSe/ZnSe/ZnS QDs with 0%, 3%, 5%, and 7% Te (Figure 3a). The EQE of the QD-LEDs was 8.9%, 9.8%, 11.6%, and 14.4% at Te concentrations of 0%, 3%, 5%, and 7%, respectively, even though all the QDs had similar PL QYs ( Figure 3b). Furthermore, the current densities of the devices increased with the Te concentration ( Figure 3d shows the impedance analysis based on the capacitance-voltage (C-V) relationships. As the carriers were injected, the capacitance increased and reached a maximum, and then, it sharply decreased as excitons recombined. The increase and decrease in the capacitance occurred faster as Te doping increased, indicating that charge injection and recombination were accelerated by the Te-induced hole states. Furthermore, the peak capacitance (built-up charges) decreased as the Te concentration increased. The photoelectron yield spectra showed that the VBM shifted from 5.44 eV to 5.35 eV as the Te concentration increased from 0-7% (Supplementary Figure 8). This difference agreed with the change in the emission peak (~128 meV). Hole-only devices (HODs) and electron-only devices (EODs) consisting of the ZnTeSe/ZnSe/ZnS QDs with various Te concentrations revealed that the hole and electron current densities at 8 V increased by 2 times as the Te concentration increased from 0-7% (Supplementary Figure 9). This suggested that the hole injection facilitated by the Te-induced low lying VBM reinforced the electron injection by the induced electric eld to lead to more e cient exciton recombination 21 . We further investigated the photophysical properties of the QDs under electrical bias. The PL spectra of QD emitting layers were measured at low voltages (< ~2.5 V) before the QD-LEDs showed electroluminescence (Figure 3c). The PL intensity of ZnSe QDs decreased the most by 20%, although the QD with 7% Te maintained the initial value under bias. In addition, the degree of the peak shift (Supplementary Figure 10) increased with the Te concentration. This could be because weakly bound holes at the valence edge of ZnSe transferred to adjacent conducting layers under the electric eld, causing easier non-radiative recombination compared to strongly bound localised Te hole states. The effective mass approximation calculations performed considering the external electric eld also veri ed that the e/h overlap of the QD with 7% Te was maintained better compared to ZnSe QDs under the electric eld (Figure 3e and 3f).
The dependency of PL on the Te concentration could be interpreted as a quantum-con ned Stark effect (QCSE), which typically causes the electric eld dependence of the peak shift and emission intensity change 22,23 . The contribution of the QCSE was examined by creating a capacitor structure (Figure 4a) to strictly allow the electric eld effect without charge ow. As the Te-doped states had a negligible effect on the spectral responses, we could con rm that the QCSE was not responsible for the spectral transitions.
Two other kinds of semicapacitor devices were fabricated to distinguish between the effects of holes ( Figure 4b) and electrons ( Figure 4c). Interestingly, the trend of the PL changes in the capacitive EOD (ITO/PMMA/QD/ZnMgO/Al) was the same as that in the LEDs, and the capacitive HOD  Table 4), and the QD lm blue-shifted from 466 nm (2.66 eV) to 459 nm (2.70 eV) as the excitation power changed from 0.01 µW to 36 µW. The EL spectra also blue-shifted from 470 nm (2.64 eV) to 461 nm (2.69 eV), and the spectral width decreased as the applied bias increased from 2.8 V to 8 V (Figure 5c). The PL and EL spectral changes with the exciton density strongly depended on the Te concentration. Furthermore, the same effect of Te doping was observed on the wavelength shift of single dot PL (Supplementary Figure  13). As PL and EL spectral behaviours were quantitatively similar, the exciton density of the QD-LEDs could be derived from the correlation obtained in the power dependent PL measurements (Supplementary Figure 14). The QD-LEDs could hold multi-excitons over ~3.7 V with a current density of 150 mA/cm 2 (dashed line in Figure 5c). As the corresponding brightness of the blue QD-LEDs with 7% Te was 17,000 cdm −2 at 3.7 V, it was presumed that the practical operation of commercial products might not require multi-exciton generation conditions. A previous study reported the spectral blue-shift in the multi-exciton region was induced by strong Coulomb repulsion between con ned holes in the biexciton state 27 .
Synthesis of ZnTeSe/ZnSe/ZnS QDs. In a 250 ml ask, 4.8 mmol of Zn(OAc) 2 and 9.6 mmol of OA were mixed with 80 mL of the TOA solvent with stirring, and the mixture was evacuated at 120°C for 15 min. Then, the mixture was heated to 280°C to prepare the Zn(OA) 2 precursor under N 2 ow (800 cc/min) and cooled to 220°C. The ZnTeSe core (optical density (OD) at 100-fold dilution was 0.54 at the rst absorption peak, 8 ml) was rapidly injected into the Zn(OA) 2   Material characterisation. The absorption and PL spectra of the QDs were measured by employing a UVvis spectrometer (Cary 5000, Agilent) and a uorescence spectrophotometer (F7000, Hitachi), respectively. The low temperature PL spectra were measured at a liquid nitrogen temperature using an equipped low temperature accessory. The PL QY was determined using an absolute PL QY spectrometer (Quantaurus-QY, Hamamatsu). The power-dependent PL spectra of the QDs were measured using a confocal microscope with a circularly polarised picosecond-pulsed excitation laser of 395 nm (LDH-D-C-390, Picoquant) with a repetition rate of 1 MHz. The laser beam was focused on the dry-type objective (UplanSApo, 0.16 NA, 4x, Olympus). The collected photons were re ected by a dielectric mirror (MRA12-E02, Thorlabs), and they entered the input slit of a dual-port monochromator (Kymera 193i-B, Andor), which was directly connected to a charge coupled device (iVac 316, Andor) and photon counting detectors (PMA hybrid 40, Picoquant). Transmission electron microscopy (TEM) analysis was performed using a Titan ChemiSTEM electron microscope operated at 200 keV. The inductively coupled plasma (ICP) atomic emission spectroscopy analysis of Zn, Se, Te, and S was performed using ICPS-8100 (Shimadzu). X-ray diffraction (XRD) patterns were recorded using a diffractometer (D8 Advance, Bruker) with a Cu-Kα source. The ionisation potential was measured by a photoelectron spectrophotometer (AC3, Riken Keiki) in air. Time-resolved absorption spectra were obtained using an ultrafast pump-probe spectroscopy system containing a Ti:sapphire regenerative ampli er (Libra, Coherent), optical parametric ampli er (TOPAS, Light conversion), and transient absorption (TA) spectrometer (Helios, Ultrafast Systems). A part of the output of the ampli er (pulse duration = 50 fs; a repetition rate = 1 KHz; centre wavelength = 800 nm) was used to pump the optical parametric ampli er that delivered optical pump pulses at 400 nm for TA measurements. The pump pulses were chopped at 500 Hz, and the probe spectra were directed to a fast CMOS array detector. All TA spectra were measured at a pump power of ca. 100 µW to suppress any undesired multi-exciton effects. All measurements were carried out in a vigorously stirred octane-dispersed QD solution.
Single-dot measurement. Octane-dispersed QDs were highly diluted with a 1 wt% PMMA toluene solution and spin coated onto a cleaned glass coverslip at a speed of 3000 rpm for 60 s. The QD-coated coverslip was encapsulated with an epoxy resin (s-209, Devcon). Single-dot measurement was performed using a laser scanning confocal microscope with a circularly polarised picosecond-pulsed excitation laser of 395 nm (LDH-D-C-390, Picoquant) with a repetition rate of 1 MHz. In the setup, a scanning galvo-mirror was used to obtain an image from different sample areas. The laser beam was focused on the oil immersion objective (UplanSApo, 1.4 NA, 100x, Olympus). The collected photons were split by a nonpolarised cube beam splitter. They entered the input slit of a polychromator (Kymera 193i, Andor), which was directly connected to an electron multiplying charge coupled device (ixon897, Andor) and the active area of a single photon avalanche photodiode (PDM series, MPD). Detected single-dot photons were registered by a time-correlated single photon counting card (TCSPC, TimeHarp 260, Picoquant), which operated in the rst-in-rst-out regime. The full width at half maximum (FWHM) of the overall instrumental response function was approximately 400-500 ps. The Symphotime 64 software (Picoquant) was used for data acquisition and processing.
Device fabrication and characterisation. Patterned indium tin oxide (ITO) was coated on glass substrates (Techno Print Co., Ltd., sheet resistance ~10 ohm/square, 2 inch × 2 inch, ITO thickness = 150 nm), which were washed with IPA in an ultrasonic bath. Then, the substrates were dried and exposed to UV-ozone treatment for 20 min (Jelight, UVO144AX-220). A poly(ethylenedioxythiophene):polystyrenesulfonate (PEDOT:PSS, Heraeus Clevios, AI 4083) solution was spin coated on the ITO glass at 2200 rpm for 50 s and annealed at 150°C for 10 min. Subsequently, the samples were transferred to a N 2 -lled glove box and annealed at 150°C for 30 min. Poly (9,9- Ab-initio calculation. The ZnSe core with a diameter of 3.6 nm consisted of 523 Zn and 555 Se atoms. The number of Te atoms in the ZnTeSe cores with 3%, 5%, and 7% Te was 17, 28, and 39, respectively. The energy levels and oscillator strength for optical transitions in the ZnTeSe cores were calculated using the pseudopotential density functional method with a plane-wave basis set. The exchange correlation of electrons was treated within the Perdew-Burke-Ernzerhof generalised gradient approximation revised for solids. The cut-off energy for the expansion of wave functions and potentials in the plane-wave basis was set as 300 eV. We used the projector augmented wave pseudopotentials provided by the Vienna ab initio simulation package. Atomic relaxation was carried out until the Hellmann-Feynman forces were less than 0.03 eV/Å. The ZnTeSe core (diameter d = 36 Å) was generated by cutting out a zincblende (ZB) bulk structure, and cluster surfaces were passivated by hydrogen atoms with charge of 0.5 or 1.5. The vacuum spacing between cluster surfaces was set as 10 Å, which was su cient for minimising arti cial interlayer interaction. The conduction band was up-shifted by 1.60 eV to ensure that the calculated ZnSe bulk band gap (1.12 eV) matched the experimental value (2.72 eV). Even though the calculated band gap was considerably smaller than the experimental band gap, the localised Te states were accurately described by the density functional theory calculations. The Coulomb interaction energies between the electrons and holes of the ZnTeSe cores (-0.140 eV--0.139 eV for x = 0-0.07) were estimated using the effective mass approximation method.
Effective-mass approximation. The HOMO-LUMO energy gaps of the QDs were calculated using effective mass approximation. The electron effective masses of ZnTe, ZnSe, ZnS used in the calculation were 0.09, 0.16, and 0.39; the respective hole effective masses were 0.6, 0.75, and 1.76; the bulk energy band gaps were 2.26, 2.72, and 3.68 eV; and the relative permittivities were 10.3, 9.1, and 8.9, respectively. The overall effective mass and relative permittivity of the ZnTeSe cores was averaged by the mole fraction of the elemental components, and the bulk energy band gap was obtained from E g = 2.72 -1.837x + 1.450x 2 . The valence band offset between the ZnTeSe cores with 3%, 5%, and 7% Te and ZnSe was calculated as 0.06 eV, 0.10 eV, and 0.14 eV, respectively. An external electric eld was applied by placing the QDs at the centre of a cylindrical vacuum space (diameter = 60 nm, length = 100 nm) in which both sides were composed of electrodes. The electric potential distribution in the QDs for various voltage differences between the electrodes was calculated by solving Laplace's equation using the nite element method implemented in the COMSOL Multiphysics software.