Recently, III-nitride deep-ultraviolet (DUV) light-emitting diodes (LEDs) have been regarded as the most suitable successors to displace mercury lamps because of their advantages in water purifying systems, medical devices, air sterilization and biological detection [1–4]. Compared to the commercial high-brightness blue LEDs, conventional DUV-LEDs still sustain poor internal quantum efficiency (IQE) [5, 6]. One important reason is that high-density threading dislocations in Al-rich materials dramatically reduce radiative recombination efficiency by acting as non-radiative recombination centers [7, 8]. The other primary factor is that the poor carrier injection results in the reduction of recombination proportion in the active region [5, 9–12]. To overcome this challenge, some designs have been proposed, e.g. optimization of quantum well (QW) width [13], staggered QWs [14], gradually increased quantum barrier (QB) thickness [11], graded QBs [15], graded-Al-composition last QB [16] and proper electron blocking layer (EBL) designs [17–20]. These methods primarily focus on designing proper active region and p-type EBL to alleviate electron leakage into the p-type layer and enhance hole injection into active region. In fact, the design of n-type layer also plays an important part in improving the performance in GaN-based blue and green LEDs. In terms of crystal growth, on one hand, the new design in the n-type region is more easily realized compared to those in the active region and p-type region. On the other hand, modulating n-type region facilitates the relaxation of stress and the improvement of crystalline quality. From the aspect of energy-band engineering, a proper design of n-type region can improve optoelectronic performance. Ali and Xing et al. reported an effective advantage of relaxing the stress and improving the blue LED quality by introducing a 200-nm n-Al_0.03Ga_0.97N underlayer and n-InGaN/GaN superlattice layer, respectively [21, 22]. Xing and Jang et al. reported that the electron capture efficiency in blue LEDs can be enhanced by introducing an n-AlGaN or graded n-InGaN layer before the growth of multiple quantum wells MQWs [23, 24]. Park et al. reported that an n-type graded-superlattice layer in green LEDs not only acts as a stress-relaxing buffer layer to release the residual stress in active region, but also acts as an electron cooler to enhance the electron capture rate in active region [25]. The above studies prove that a proper design of n-type region has great advantage in improving the crystalline quality and optoelectronic performance of blue and green LEDs. However, there are few reports on the design of n-type layer for DUV-LEDs.

In this study, a novel DUV-LED structure with an n-type electron deceleration layer (EDL) is proposed. This structure can effectively modulate the electron distribution by increasing the electron capture efficiency. Simultaneously, it can pull up the valence band of MQWs to improve hole transport. Through such an energy-band engineering, DUV-LEDs with improved radiative recombination rate and IQE were achieved.

Four DUV LED structures (Figure 1) were designed, denoted as the Original, D1, D2, and D3. The optoelectronic performances of DUV LEDs were simulated using the Advanced Physical Model of Semiconductor Devices package from the Crosslight software, based on a Schrödinger-Poisson solver method. The Original LED was designed on a c-plane sapphire substrate, consisting of a 2.5-um-thick AlN template, a 3-µm-thick n-Al_0.55Ga_0.45N layer (electron concentration: 6 × 10¹⁸ cm⁻³), 5-pair 2-nm-thick Al_0.45Ga_0.55N/10-nm-thick Al_0.55Ga_0.45N MQWs, followed by a 10-nm-thick p-Al_0.60Ga_0.40N EBL (hole concentration: 3 × 10¹⁸ cm⁻³). Finally, the structure was capped by a 10-nm-thick p-Al_0.55Ga_0.45N layer (hole concentration: 5 × 0¹⁷ cm⁻³) and followed by a 100-nm-thick p-GaN (hole concentration: 1 × 10¹⁸ cm⁻³) layer. D1-D3 have the same design with the original LED except for the EDL. The D1 EDL is composed of Al_0.48Ga_0.52N (12.0 nm). The D2 EDL is composed of Al_0.56Ga_0.44N (4.0 nm)/Al_0.48Ga_0.52N (4.0 nm)/Al_0.56Ga_0.44N (4.0 nm). The D3 EDL is composed of Al_0.56Ga_0.44N (2.4 nm)/Al_0.52Ga_0.48N (2.4 nm)/Al_0.48Ga_0.52N (2.4 nm)/Al_0.52Ga_0.48N (2.4 nm)/Al_0.56 Ga_0.44N (2.4 nm).

The Shockley-Read-Hall lifetime was chosen to be 10.0 ns in the simulations [26]. The internal background loss was set as 2000 m⁻¹ [27, 28]. Because of the polarization effect, the polarization charge of the built-in interface calculated by Fiorentini et al. was reduced by a factor of 0.4 [10, 29]. The Auger recombination coefficient was set to 1 × 10^–30 cm⁶ s⁻¹, and the operating temperature was 300 K [5]. The energy band-offset ratio at the AlGaN/AlGaN heterojunctions was set to 0.7/0.3 [30]. The other model parameters used in the simulation can be referenced [31, 32].

The IQE and LOP curves concerning the driven current are plotted in Figures 2A, B, respectively. From Figure 2A, we can see that D3 obtains the maximum IQE of 41.74%, however, the IQEs of the original device, D1 and D2 are 29.57%, 32.96% and 36.81%, respectively. Figure 2B shows that the LOPs of the four devices increase linearly with the current injection. The LOPs of the Original, D1, D2 and D3 LEDs are 24.60, 27.21, 30.26, and 34.23 mW at 200 mA, respectively. Obviously, compared to the original LED, D3 exhibits a superior LOP with an improved coefficient of 1.39.

These results suggest that D3 achieved the highest IQE and LOP due to the introduction of a proper EDL.

Figure 3 depicts the energy band diagrams of four LEDs to exhibit the origin of enhanced optoelectronic performance. The green region represents the EDL area. ΔE_c1 is defined as the maximum potential barrier height for an electron in the front of the first QB. It can be seen that the values of ΔE_c1 are quite different due to the different spontaneous and piezoelectric polarization caused by the inserted EDLs. The calculated value of ΔE_c1 for the original device is only 32.60 meV. By inserting the new EDL layers, the values of ΔE_c1 for D1, D2, and D3 significantly increase to 155.90, 156.19, and 157.99 meV, respectively. ΔE_c2 is defined as the first potential barrier height for electrons in the MQWs along the growth direction. Since the four devices have the same structure parameter in the first QB, the values of ΔE_c2 do not change much. The calculated values of ΔE_c2 for the original device, D1, D2, and D3 are 117.97, 119.76, 118.01 and 115.91 meV, respectively. The electron injected from n-type layer into MQWs will go across ΔE_c1 and ΔE_c2 successively. Therefore, the electron blocking effect depends on the combined action of ΔE_c1 and ΔE_c2. Obviously, compared with the electron blocking effect of the original device, that of D3 is significantly enhanced, which decelerates the velocity of electrons. Besides, the electrons have also been stored in EDL (depicted by an orange arrow), further decelerating the velocity of electrons. To understand the effect of the inserted EDL on the holes in the active region, we take the middle QW as an example. The hole potential heights ΔE_v of these four structures are 152.98, 151.04, 149.42, and 147.72 meV, respectively. In this regard, the new design can also pull up the valence band of QBs to promote hole transport. Compared to the original LED, the LEDs with EDL not only present higher barrier heights for electrons, but also lower barrier heights for holes. Thus, the electron-hole injection efficiency can be simultaneously improved via effective structural designs.

Figures 4A, B depict the electron and hole concentration distributions in these LEDs at 200 mA. The electron and hole concentrations of D1, D2, and D3 MQWs are all higher than those of the original device. As calculated, the electron concentration of D3 increased by 18.25% compared to the original device. At the same time, the hole concentration of D3 is higher than that of the original LED by 22.01%. Figure 4C displays the calculated electrostatic fields of the four structures to investigate the carrier’s radiative recombination in the MQWs. The electrostatic fields exist due to the strong polarization induced by the lattice mismatch. It can be observed that the electrostatic field of the original device is higher than those of D1-D3. The existence of strong electrostatic field in the original device deteriorated IQE by decreasing wave function overlap of electrons and holes, which is known as Quantum-confinement Stark Effect (QCSE). The lower electrostatic field in D1-D3 LEDs helps to increase the overlap of electron-hole wave functions, and then it can improve the radiative recombination efficiency.

Figure 5A demonstrates the radiative recombination rates in the active region of the four structures at 200 mA. D3 shows a much higher radiative recombination rate than the original LED. The improved radiative recombination rate can be attributed to the increase of carrier concentration and the decrease of the electric field in the active region. It is consistent with the previously mentioned energy band diagrams, carrier distribution, and electric field. Figure 5B shows the spontaneous emission spectra of four designed structures at an injection current of 200 mA. The spectral peak intensity of the D3 is nearly 1.59 times that of the original LED. At the same time, the peak emission wavelengths of the four samples are at 288 nm, indicating that the new structural design has no influence on the wavelength.

We have numerically analyzed the impact of an n-type EDL on the energy-band engineering for DUV-LEDs. The optoelectronic performances of the DUV-LEDs were remarkably improved, including the radiative recombination efficiency, IQE, and LOP. Introducing an EDL not only increases the potential barrier height for the electrons but also pulls up the QB valence band, which significantly modulates the carrier distribution. This can improve CIE and the radiative recombination rate, compared to the original structure. Therefore, the proposed device architectures in this study should help the community to fabricate high-efficiency DUV-LEDs.