Deep ultraviolet (DUV) light has various applications in the areas of air/water purification, disinfection, sterilization, and free-space optical communication^[1–3]. As an emerging clean DUV light source, AlGaN-based DUV light-emitting diodes (LEDs) have been recognized as a promising candidate to substitute the toxic mercury-based UV lamps. However, the state-of-the-art DUV LEDs still exhibit low light output power (LOP) and poor wall-plug efficiency (WPE)^[4]. They are mainly caused by large densities of dislocations/defects^[5], the inefficient p-type doping in the Al-rich AlGaN alloys^[6], the existence of strong quantum-confined Stark effect^[7], and the low light extraction efficiency in the devices^[8]. Many efforts have been devoted to addressing the challenges mentioned earlier. Exciting progress has been made to improve the optical performance of the DUV LEDs by either employing novel device structures^[9–13] or adopting new growth/epitaxy strategies^[14–17].

Recently, the implementation of the tunnel junction (TJ) structure has gained remarkable attention in the development of high-efficiency group-III-nitride-based LEDs. In one aspect, the high-resistance p-type AlGaN contact layer is replaced by a relatively low-resistance AlGaN-based TJ. Therefore, the current injection efficiency can be significantly improved in the DUV LEDs^[18,19]. In another aspect, the TJ has been used to construct visible cascaded LEDs^[20–24]. Unlike the conventional LED structure with a single active region^[25,26], stacked multiple active regions can significantly enhance the optical power due to the repeated electron–hole (e–h) recombination process. So far, such a cascaded structure has not yet been explored in the development of high-efficiency AlGaN-based DUV LEDs. In this work, we tentatively construct a unique DUV LED by cascading two DUV active regions interconnected via a specially designed n+-AlGaN/p+-AlGaN TJ. Here, we name it TJ-cascaded (TJC) DUV LED. We found that the output power and WPE of the proposed TJC-DUV LED can be considerably boosted through multiple recombination processes in the cascaded active regions. Then, the underlying physical mechanism of performance enhancement is thoroughly discussed and theoretically analyzed. We believe that the TJC-DUV LED design not only reduces the cost of the fabrication and packaging, but also miniaturizes the LED chip size through vertical integration of two “LEDs” in the pursuit of high-efficiency DUV LEDs of the future.

In this study, three devices are constructed and investigated. Firstly, a conventional DUV LED (C-DUV LED) emitting at 284.5 nm, experimentally reported by Yan et al.^[27], was utilized as a reference LED structure. Our physical model and simulation parameters were strictly calibrated to fit our simulated curves with their experimental data to validate our model. As shown in Fig. 1(a), the reference LED structure is composed of a 3-µm-thick n-Al0.6Ga0.4N layer (n-type doping concentration of 5 × 10¹⁸ cm⁻³), five 3-nm-thick Al0.4Ga0.6N quantum wells (QWs) embedded within six 12-nm-thick Al0.5Ga0.5N quantum barriers (QBs), 20-nm-thick p-Al0.65Ga0.35N electron-blocking layer (EBL), 50-nm-thick p-Al0.5Ga0.5N (p-type doping concentration of 1×1019cm−3), and 120-nm-thick p+-GaN (p-type doping concentration of 1×1020cm−3). Secondly, we also used the same C-DUV LED but with ten QWs in the active region for comparison. Lastly, the TJC-DUV LED consisting of two active regions (each active region has five QWs), which are bridged with a TJ (10-nm-thick p+-Al0.5Ga0.5N/10nm-thick n+-Al0.5Ga0.5N), is designed, as shown in Fig. 1(b). The n-type and p-type doping concentrations in the TJ are both 1×1020cm−3, according to the reported experimental values^[28,29]. It should be noted that the thickness of the n-Al0.6Ga0.4N layer (n-type doping concentration of 5×1018cm−3) for the TJC-DUV LED is thinned to 50 nm for reducing the series resistance of the device. The second active region is built on the 50-nm-thick n-Al0.6Ga0.4N layer with the same parameters, followed by the p-AlGaN layers with the same doping and thickness.

The simulator is provided by Crosslight Inc. APSYS (version 2018), which is widely used in the investigation of the III-nitride-based LEDs in both academia and industry^[30]. The Shockley–Read–Hall (SRH) recombination lifetime, the Auger coefficient, and the light extraction efficiency are set as 15 ns, 2.88×10−30cm6/s, and 13%, respectively. The band offset ratio is set to be 0.5/0.5 for the III-nitride system to fit the power-current (L-I) curves and current-voltage (I-V) curves, which are extracted from the experimental results reported by Yan et al.^[27]. The polarization-induced charge density is assumed to be 50%, considering the screening effect of defects in the epitaxial layers^[31]. Other material parameters used in this work can be found in Ref. [32]. Regarding the tunneling model implemented in this work, we follow the classical model and derivation, which are explained in Ref. [33]. The e–h pair generation rate from band-to-band tunneling in the reverse-biased TJ can be calculated as the following^[34]: G(F)=(qFm*)(2π2h3)P0E∥,(1)where q is the fundamental electron charge, F is the local field, m* is the effective tunneling mass, and h is Planck’s constant. P0 is the tunneling probability with a momentum of zero perpendicular. E∥ is the electron kinetic energy in the tunneling direction. This expression for a local field-dependent generation rate can be subsequently substituted into the drift-diffusion equation solver as a carrier generation term during the III-nitride-based LED operation^[14,15].

The optical performance of the investigated three DUV LEDs is exhibited in Figs. 2(a)–2(d). We found that the simulated results [blue curves in Figs. 2(a) and 2(d)] perfectly matched with those experimental results extracted from the reference LED (orange squares from Yan et al.^[27]), suggesting the high reliability of the physical parameters and models implemented in this simulation process. As seen in Fig. 2(a), the TJC-DUV LED has an LOP of 14.4 mW at 60 mA, whereas the C-DUV LED devices delivered only 6.7 mW (5 QWs) and 6.2 mW (10 QWs), respectively. In other words, the LOP of the TJC-DUV LED is 114.9% higher than that of the C-DUV LED with five QWs under the same injection current of 60 mA, attributing to a much higher internal quantum efficiency (IQE) of the TJC-LED, which will be discussed later. The great advantages of the TJC-DUV LED can be further confirmed in Fig. 2(b), which shows the electroluminescence (EL) spectra of the C-DUV LED and the TJC-DUV LED. It can be observed that the peak intensity of the TJC-DUV LED shows the highest value, which is consistent with the LOP performance in Fig. 2(a). When introducing a TJC structure, the luminescence intensity is more than doubled, indicating the successful operation of the TJC-DUV LED device.

There is one question to be answered: can we simply increase the number of QWs to boost the optical performance instead of using a complex TJC structure? In other words, it might not be necessary to build a TJC structure to enhance the performance of the DUV LED device. To address this issue, we investigated the LOP and IQE performance of a C-DUV LED by increasing its number of QWs from 1 to 10. The results are shown in Fig. 2(c). When the QW number increased from 1 to 4, the LOP and IQE of the C-DUV LED can be enhanced dramatically, attributed to the decreased electron current overflow^[35]. While if we continuously increase the number of QWs from 5 to 10, the LOP and IQE values decreased abruptly, mainly caused by the limited hole injection efficiency. Therefore, we conclude that it might not be feasible to enhance the optical performance by simply increasing the number of QWs in the C-DUV LED. Instead, the DUV LED performance can be significantly boosted after introducing a TJC structure, as we proposed in this work [Figs. 2(a) and 2(b)]. However, it is certain that the turn-on voltage of the TJC-DUV LED with two interconnected active regions is larger than that of the C-DUV LED with five QWs or 10 QWs [in Fig. 2(d)]. This phenomenon has also been observed in the cascaded visible LEDs previously^[13–17] and can be well explained by the doubled built-in electric fields of stacked two active regions in the TJC-LED structures.

The output power of the TJC-DUV LED is much larger than that of the C-DUV LED with the same power supply, as compared in Fig. 3(a). The WPEs for C-DUV LEDs and a TJC-DUV LED are demonstrated in Fig. 3(b). The C-DUV LED with 10 QWs exhibits a lowered WPE than that with five QWs due to the trade-off between the number of QWs and carrier concentration in the active region. In comparison, the WPE of the TJC-DUV LED is significantly increased, resulting from the reduced injection current level at the same value of LOP among the three devices. It can be noted that the WPE of the TJC-DUV LED is a little smaller at a low current level, but it has a much larger value at a high current level than that of the C-DUV LED with five QWs. To be more specific, the WPE of the TJC-DUV LED is significantly boosted by 25% at 60 mA than that of the C-DUV LED with five QWs. It is mainly attributed to the increasing operation bias of the TJC-DUV LED compared with the bias of the C-DUV LED with five QWs.

Figure 4 represents the band diagrams of the TJC-DUV LED. In the C-DUV LED, the recombination of the electrons and holes takes place only once within the active region to generate photons. In contrast, the photon generation process can occur in both active regions in the TJC-DUV LED, enabling a dual radiative recombination process, as illustrated in Fig. 4. A mode of smooth carrier transport and injection can be provided by the TJ structure, which bridged the two active regions. Specifically, the electrons in the p side move inside the valence band to the TJ, which transfers these valence band electrons into the conduction band of the second active region. In the meantime, the holes in the p side of the TJ move in the opposite direction until they reach the first active region. Finally, the photon generation process can be repeated in the second active region of the TJC LED.

The e–h concentrations and radiative recombination rate profiles in the active regions are further investigated to explore the underlying mechanism of the performance enhancement for the TJC-DUV LED. As illustrated in Figs. 5(a1) and 5(b1), the e–h concentration in each QW is drastically reduced when the number of QWs is increased from 5 to 10 in the C-DUV LED. Consequently, the C-DUV LED with 10 QWs exhibits lower radiative recombination rates within the active region in comparison with the C-DUV LED with five QWs, as exhibited in Figs. 5(a2) and 5(b2). On the contrary, the e–h concentration within the two active regions (10 QWs in total) in the TJC-DUV LED remains at the same level as those in the C-DUV LED with five QWs, according to Figs. 5(a1) and 5(c1). The enhanced e–h concentration within the two active regions of the TJC-DUV LED can be ascribed to the dual injection of the electrons and holes in two PIN junctions connected via a TJC structure. Besides, the multiple radiative recombination behavior is also observed in both two active regions, as illustrated in Fig. 5(c2). As shown in Figs. 5(d) and 5(e), the non-radiative recombination rates of the TJC-DUV LED are obviously lower in each QW than that of the C-DUV LED. Hence, with the same radiative rates, the TJC-DUV LEDs possess a significantly higher IQE, as presented in Fig. 5(f), which is attributed to the LOP enhancement of the TJC-DUV LED rather than that of the single C-DUV LED by more than 100%.

In summary, we have numerically investigated the unique TJC-DUV LED with cascaded active regions connected by a TJ. The simulation results show that introducing the TJC structure in the DUV LEDs can successfully enable the repeated use of electrons and holes for radiative recombination and consequently produce a significantly enhanced LOP and WPE of the device. The proposed TJC structure provides a reliable and cost-effective method to achieve high-performance DUV LEDs, even possibly in high-power DUV lasers of the future.