Deep-ultraviolet (DUV) light-emitting diodes (LEDs) have emerged as promising light sources with a wide range of potential applications, including air-water purification, surface disinfection, bio-medical treatment, UV curing, analytical science, and security [1,2]. The advantages of AlGaN-based DUV LEDs such as the compact module system, low working voltage, fast operation speed, and environment friendly materials render them feasible substitutes for conventional high-pressure Hg-lamps in various application scenarios. In addition, this solid state DUV source has been developed to support other innovative applications [3], such as three-terminal light-emitting and detection devices, demonstrating great potential in advanced optoelectronic integrated circuits. For DUV LEDs with emission wavelength shorter than 250 nm, the much higher photon energy enables more effective disinfection of multiresistant microorganisms and decomposition of organic waste in water [4,5]. However, the current power conversion efficiency (PCE) or wall-plug efficiency (WPE) for the sub-250 nm DUV LEDs remains low due to the following three main challenges. First, the growth of high-quality AlGaN templates for AlGaN quantum wells (QWs) with low defect densities is difficult. The threading dislocation density in AlGaN QWs was about ${10}^9{\mathrm{cm}}^{-2}$, which is comparatively high compared to InGaN/GaN-based materials, leading to unexpectedly high non-radiative recombination in the devices [6]. Second, the light extraction efficiency (LEE) for high-Al-fraction AlGaN quantum structures is low because of the dominant transverse magnetic (TM) mode emission induced by valence band splitting inversion and the small critical angle for light propagation to free space [7]. Although various field manipulation techniques have been proposed to enhance the IQE of AlGaN QWs [8,9], effective control of the TM mode emission remains limited. Third, achieving effective p-type doping within the high-aluminum (Al)-fraction p-AlGaN contact layer is a challenge, as the activation energy ($E_A$) of p-type doped Mg atoms gradually rises and the solubility of Mg atoms in AlGaN decreases with the increase in Al content [10]. As a compromise, p-GaN is generally used as the contact layer, but its narrow bandgap induces severe absorption of light emitted towards the p-layer. Among these challenges, the LEE primarily limits the final WPE to less than 1% in sub-250 nm DUV LEDs [1]. Thus, it is urgent to improve the LEE.

There have been different strategies to improve the LEE of DUV LEDs, such as optimized device structure, highly reflective metal electrodes, patterned substrate, nanostructured surface, and various plasmonic mediums, to manipulate the light transmission and suppress the absorption loss. Yu et al. reported that micro-structured DUV LEDs with smaller chip areas could deliver significantly higher light output power density [11]. Furthermore, the rationally designed chip sidewall angle could further improve the light extraction originating from the enhanced reflection of the inclined sidewall [12,13]. The highly reflective p-type electrode also has proven to be an effective strategy to improve the total power output for the flip-chip DUV LEDs, such as using the highly reflective photonic crystal [14], Ni/Al (Mg) or Rh p-type electrodes [15–17], and ${\mathrm{SiO}}_2$ microcavity incorporated Al reflector [18]. In addition, the traditionally used nano-patterned substrate or AlN template has been intensively investigated to boost the light output for the DUV quantum structures, which benefited from the simultaneous improved crystal quality of epitaxial layers by epitaxial lateral overgrowth and the manipulated light transmission by the periodically modulated dielectric environment [19–21]. To reduce the light loss caused by internal total reflection on the light-emitting side of the flip-chip device, various surface microstructures have been introduced into DUV LEDs, including nanocolumns, nanocones, nanowire arrays [22–24], etc. As the surface plasmon resonance (SPR) could couple with the excitons to improve the recombination of carriers or act as a scattering medium to change the transmission mode of photons [25–27], metallic nanostructures also have been developed to improve the light extraction. However, it should be noted that the current LEE for the DUV LEDs was still lower than 25%, especially for the devices with much shorter wavelengths below 250 nm [28]. The main bottleneck restrictions are the difficulty in the fabrication of the transparent p-AlGaN contact epilayer and the intrinsic strong TM mode emission for the high-Al-content AlGaN quantum wells, which seriously limited most of the above strategies utilized in the devices.

In this work, we demonstrate enhanced light extraction in sub-250 nm DUV LEDs through a novel flip-chip device structure incorporating uniformly distributed n-type contact rods and embedded plasmonic omni-directional reflective pad arrays in the p-GaN layer. The distributed n-contact rods that cross the p-GaN layer not only ensure uniform current spreading but also simultaneously act as efficient light-guiding channels to facilitate photon extraction. The Al-based plasmonic pads embedded within the p-GaN layer proficiently manipulate photon wavevectors via plasmon coupling, while exerting minimal interference on hole transport within the p-GaN layer. Consequently, the omni-directional reflective mechanism through the plasmonic pads led to a significant 12.5% increase in light output in contrast to conventional devices lacking such structures. Devices with optimized plasmonic pad size achieved an average optical output power of 2.34 mW at 100 mA, which was further enhanced to 3.45 mW through integration with a fluorine-resin-bonded quartz lens. This study not only demonstrates the efficacy of embedded omni-directional plasmonic reflective nanostructures in controlling photon propagation and improving LEE in sub-250 nm AlGaN LEDs but also presents a viable solution for developing high-performance DUV LEDs while addressing the fundamental challenge of achieving effective p-type doping in high-Al-content AlGaN materials.

The epitaxial structure for the AlGaN-based DUV LED with center wavelength at $\sim 249\mathrm{nm}$ was grown by metal-organic chemical vapor deposition (MOCVD) on the regular flat sapphire substrate with a thickness of 1000 μm. The wafer was composed epitaxial layers from bottom to top of a 3.2-μm-thick AlN template layer, a 0.8-μm-thick unintentional doped $\mathrm{n}\text{-}{\mathrm{Al}}_{0.8}{\mathrm{Ga}}_{0.2}\mathrm{N}$ layer, a 0.9-μm-thick Si-doped $\mathrm{n}\text{-}{\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}$ layer, five pairs of ${\mathrm{Al}}_{0.70}{\mathrm{Ga}}_{0.30}\mathrm{N}/{\mathrm{Al}}_{0.75}{\mathrm{Ga}}_{0.25}\mathrm{N}$ (2.5 nm/12.5 nm) multiple quantum well (MQW) active layers, a 50 nm Mg-doped p-AlGaN electron blocking layer, and a 300-nm-thick p-GaN contact layer. The detailed fabrication process of the LED chips is illustrated in Fig. 8 (Appendix A). A distributed n-contact rod (diameter of 39 μm) passing through the p-type contact layer was used to improve the uniformity of current spreading. The Al-based plasmonic reflective pad arrays were embedded in the p-GaN layer by lithography and ICP etching, followed by the depositing of a passivating ${\mathrm{SiO}}_2$ layer and Al film. A 20-nm Ni/Au layer was deposited on the p-GaN to form the ohmic contact, while the Ti/Al/Ti/Au layer was deposited in the n-contact area to form the ohmic contact. Figures 9(a) and 9(b) (Appendix A) present optical microscope (OM) images of the electrode side and sapphire side of the fabricated flip-chip device, respectively, providing a clear depiction of the distribution of the n-contact rods and plasmonic pads. For the encapsulated LED with a quartz lens, the devices and the lens were first treated by a silane coupling agent. After drying in the air, the LED chips were dipped with a drop of 20 μL fluorine resin (FM160, Shanghai FluoroLuster Materials Co., Ltd., China) solution (10% mass fraction) on the surface (sapphire substrate side), and then bonded with the quartz lens followed by the heating and curing process under a vacuum oven and a certain pressure.

A commercial electrical and optical simulation package for LED devices by SimuLED software (Suzhou STR Software Technology Co., Ltd., China) was used for the evaluation of the electrical and optical performance of the designed DUV LED, which incorporates three functional sub-modules. The quantum structure material parameters were set according to the epitaxial layers used in the experiment. The key physical parameters, such as thermal conductivity, carrier mobility, and activation energy of the dopant, were derived from the literature [29]. The remaining parameters, such as the optical properties, were obtained directly from the software library. The electrical characters for the 1D LED quantum structure were simulated first by the SiLENSe (Version 6.5.1, STR Software) module. Then, the 1D results were input to the 3D optoelectronic analysis module called SpeCLED (Version 6.5.1, STR Software) for revealing the light emission feature. Next, the ray tracing module of RATRO (Version 6.5.1, STR Software) was used for light extraction analyses. A total of $1\times {10}^7$ rays were used in the ray tracing simulation and the power detector was placed above the light emission direction of LED. The local-field distribution and LEE performance manipulated by the plasmonic pads were evaluated by the FDTD simulation method using a commercial full wave FDTD simulation package (FDTD Solutions, Lumerical), and the dipole source with a TE or TM setup was adopted as the excitation. The refractive index parameters for materials such as ${\mathrm{Al}}_x{\mathrm{Ga}}_{1-x}\mathrm{N}$, ${\mathrm{SiO}}_2$, and Al used in the simulation were obtained from literature [30,31]. The angle-dependent reflective properties for the plasmonic pads were simulated using a plane wave source with different incident angles in the FDTD simulation model.

The current-voltage ($I\text{-}V$) and light output power (LOP) curves for the DUV LEDs were measured by the HAAS-2000 high-accuracy array spectroradiometer (EVERFINE, China). And the corresponding angular emission pattern for the devices was characterized by a LEDGON 100 goniophotometer equipped with a UV spectrometer (Instrument Systems GmbH, Germany). The 2D surface near-field EL emission patterns for LEDs were recorded using a near-field optical distribution test system (YQ124) at Gold Medal Analytical & Testing Group (GMATG, Guangzhou, China). The TE and TM mode edge-emissions of the DUV LEDs were collected by a NOVA high-sensitivity spectrometer (Ideaoptics, China) through the focusing lens and Glan-Taylor prism.

Here, the AlGaN-based DUV LED structure with a center emission wavelength of $\sim 249\mathrm{nm}$ was grown by metal-organic chemical vapor deposition (MOCVD) on a sapphire substrate. A mature p-type GaN layer with a Mg doping concentration of $2\times {10}^{19}{\mathrm{cm}}^{-3}$ was employed to ensure reliable p-type contact and hole injection. The LEDs were designed to have a uniform current spreading with distributed n-contact rods (diameter of 39 μm) that penetrated through the p-electrode and were fabricated to have a chip area of $30\mathrm{mils}\times 30\mathrm{mils}$, as illustrated in Fig. 1(a). The distributed n-rods additionally act as light guiding channels to improve the light extraction, especially for the TM mode photons by utilizing the scattering effect of the metal rods with a dielectric isolation layer, as shown in Fig. 1(b). It should be noted that the device’s central region has been designed to exclude functional features (Fig. 9 in Appendix A), such as the n-contact rods, to avoid additional film steps and undulations. This reduces the risk of damage from uneven mechanical stress at the center area during fabrication, such as the chip transfer process. A 20-nm Ni/Au layer was deposited on the p-GaN to form the ohmic contact followed by the Cr/Al/Ti/Au composite metal electrodes. To further manipulate the photons’ wavevector towards high extraction efficiency, uniformly distributed Al plasmonic reflective pads with different diameters (8, 16, 20, 24 μm) and area ratios were embedded in the p-GaN layer by a lithography, ICP etching, and film deposition process, as illustrated in Figs. 1(a) and 1(b) and Fig. 8 (Appendix A).

The device performance was initially evaluated using 1D simulation with SiLENSe, employing an isothermal drift-diffusion model with quantum-mechanical corrections via quantum potential [32], followed by 3D optoelectronic analysis using SpeCLED in SimuLED software. The simulation successfully predicted DUV electroluminescence (EL) emission centered at approximately 249 nm for the quantum structures (Fig. 10 in Appendix A). These emission characteristics were subsequently imported into the RATRO ray tracing module for light extraction analysis. For the Plasmonic sample model, a configuration with 8 μm diameter holes etched in the p-GaN layer was chosen for investigation. The area ratio of plasmonic pads ($R_p$) in the LED chip was defined as the etched p-GaN area for plasmonic pad embedding relative to the total active area. Based on this definition, the $R_p$ was 0.68% for the 8-μm Plasmonic sample. $I\text{-}V$ characteristics were compared between device models with and without plasmonic pads [Fig. 1(c)]. The nearly identical $I\text{-}V$ curves demonstrate that p-GaN layer etching and plasmonic pad embedding have a negligible impact on current injection. Further simulations with larger plasmonic pad sizes yielded similar results, showing consistent $I\text{-}V$ characteristics across all configurations (Fig. 11 in Appendix A).

To reveal the detailed effect of microstructure on the spatial current spreading and injection on the LED devices, the current distribution within the active layer was extracted for comparison and shown in Fig. 1(d) for the Control and Plasmonic samples. The results showed that the distributed n-contact rods facilitated a relatively uniform current density distribution in the active region of the flip-chip device, with low standard deviations for both the Control ($30.47\pm 5.66\mathrm{A}{\mathrm{cm}}^{-2}$) and Plasmonic ($30.73\pm 5.67\mathrm{A}{\mathrm{cm}}^{-2}$) samples. The formed mesh structure by etching the p-GaN has little influence on the local current spreading (indicated by the current flow arrows) for the Plasmonic sample but with slightly enhanced average current density (indicated by the color intensity) within the active region of the device [Fig. 1(d)]. Undoubtedly, this increased current density is caused by the reduction of the p-type electrical injection area due to the etching of the p-GaN. As the size of the plasmonic pad increased, the current density also gradually increased correspondingly as the etching area ratio increased, as shown in Fig. 12 (Appendix A) and Fig. 1(e).

The light emission and extraction performance was subsequently evaluated by ray tracing on the device model. The near-field light emission pattern shown in Fig. 1(f) and Fig. 13 (Appendix A) demonstrated that the distributed n-contact rods and the plasmonic pads in p-GaN both have modulated the light propagation and that escaping from the device, shown as the enhanced light intensity near the rods and pads. This possibly originates from the enhanced scattering for the TE or TM mode photons from the active layer thus improving the light extraction. The overall enhanced light extraction of the LED device containing the plasmonic pad also can be visually seen from the far-field emission patterns (Fig. 14 in Appendix A). The extracted LEE has been improved from 10.2% to 10.5% for the Plasmonic sample as the size of the plasmonic pad increased from 0 (Control) to 24 μm [Fig. 1(g)]. It was because the LEDs were simulated under the same driving current (100 mA) and almost the same voltage, and thus the increased LEE here came from the manipulating effect of the plasmonic pads on photon transmission. Furthermore, the microscope manipulation mechanism for the light could not be considered when using the ray tracing method, and therefore an additional enhancement by the plasmonic pad should exist and this will be discussed later.

The DUV chips were subsequently fabricated in the flip-chip architecture as illustrated in Fig. 1(a). To ensure the efficient light extraction of the DUV chips, the sapphire substrate has been polished with a reduced thickness of about 300 μm. Before depositing the reflective Al films in the etched p-GaN holes, a ${\mathrm{SiO}}_2$ isolating layer with a thickness of 1 μm was deposited to passivate the defects induced by ICP etching. The devices with different diameters of plasmonic pads ($D=8$, 16, 20, and 24 μm) were fabricated on the same epitaxial wafer using the composite photolithography mask. The etched holes were uniformly distributed within the p-GaN and the area ratio ($R_p$) of the plasmonic pads is summarized in Table 1. Figure 2(a) shows the typical EL spectra for the DUV LED device with different sizes of plasmonic pads. The emission peak was located at 249.5 nm and obvious enhancement has been achieved for the Plasmonic samples when compared with the Control one. The typical emission image is shown in the inset of Fig. 2(a). A batch of the DUV LED devices was measured using the integrating sphere system to obtain the optical output power when driven at a constant current of 100 mA. The results are statically shown in Fig. 2(b) and Table 1, indicating that the average optical power of 2.08, 2.20, 2.26, 2.34, and 2.24 mW was realized for the Control sample and Plasmonic sample with diameters of 8, 16, 20, and 24 μm, respectively. As the size of the plasmonic pad increases, the optical power of DUV LED devices gradually increases and then decreases, and the maximum enhancement of about 12.5% was realized on the Plasmonic sample with a diameter of 20 μm.Table 1.

Parameters and Optoelectronic Performance of the DUV LEDs

The $I\text{-}V$ and LOP curves for the typical Control and Plasmonic DUV LEDs are shown in Fig. 2(c). As seen from the $I\text{-}V$ curves, even with the etching of holes in the p-GaN layer and the addition of plasmonic pads, the operation voltage in the Plasmonic sample does not increase compared to the Control sample. This infers that the etching of the p-GaN layer does not significantly impact the current injection, which is consistent with the previous simulation results shown in Fig. 1(c). The LOP curves follow the same trend as the previously mentioned EL spectra and the optical power of the DUV devices, with the highest LOP achieved when the plasmonic pad size is 20 μm in diameter. In addition, the LOP curves showed that the devices’ optical output power gradually decreases after the current exceeds 250 mA. This is likely due to the carrier overflow effects caused by the polarization field within the AlGaN quantum wells at high currents, especially for epitaxial structures with a wavelength shorter than 250 nm, which have a lower quantum-well-barrier energy offset [33,34]. The angular emission patterns [Fig. 2(d)] had a similar increasing trend. The full-angle emission enhancement also confirmed that the plasmonic pads introduced in this work had the omni-directional ability to make the photons escape from the device easier.

To reveal the detailed enhancement mechanism of the plasmonic pads on the device, we further measured the device’s near-field radiation pattern, as shown in Figs. 3(a) and 3(b). It can be observed from the figure that the conventional Control sample without the introduction of the plasmonic pad had a near-field light intensity distribution with strong edges distribution, with some enhancement near the n-contact rod area. This emission characteristic was consistent with the short-wavelength DUV LEDs: due to the strong TM mode emission and the severe total internal reflection from the sapphire substrate to the air medium, photons are difficult to emit directly outside the front surface of the sapphire substrate, while lots of photons are emitted from the edge area of the chip. The distributed n-contact rod structure designed in this work allowed a significant enhancement of photon emission through the scattering effect around the n-type contact rods, consistent with the structure simulated by the ray tracing simulation (Fig. 13 in Appendix A). The reflection and photon outcoupling effects of the Al-based plasmonic pads have modulated the photons’ transmission mode of TE and TM when holes are etched in the p-GaN layer and plasmonic pads are embedded, thereby significantly enhancing the photon emission efficiency. With the increasing size of the plasmonic pad, the surface light intensity of the device gradually increased, especially for the sample with a plasmonic pad size of 20 μm. For ease of analysis, the local near-field manifested images of typical samples are displayed in Figs. 3(c)–3(e), with comparison to the corresponding OM images of the device. The results showed that, for the Control sample, strong light emission was only observed around the n-type contact rod on the surface of the device, while other active regions showed weaker light intensity. A significant emission enhancement was observed around the pads with the introduction of appropriate plasmonic pads ($D=20\mathrm{\mu m}$), even though the pad itself emitted hardly any light due to the lack of p-GaN and effective hole injection. Certainly, there was a decrease in the overall intensity of the near-field distribution when the plasmonic pad was too large ($D=24\mathrm{\mu m}$), which was due to a larger sacrifice of effective light-emitting area. For all chips, the observed higher emission intensity at the center may result from the special configuration of the n-contact rods, with their symmetric arrangement leaving the center unobstructed, allowing scattered light from surrounding rods to emit from the center.

The above results confirmed that introducing plasmonic pads within the p-type layer could effectively manipulate the propagation of photons from quantum wells, thereby significantly enhancing the LEE. The near-field simulations further confirmed this manipulation characteristic, as shown in Fig. 4. Without the introduction of plasmonic pads, both TM and TE mode dipole sources encounter substantial optical loss due to absorption in the p-GaN, as depicted in Fig. 4(a). The near-field distribution extracted from FDTD simulations showed that only a portion of photons emitted by the dipole source could successfully escape towards the sapphire side when below the critical angle, while a large number of photons encounter absorption in p-GaN during propagation [Figs. 4(b) and 4(c)]. The photons’ transmission pathways for both TE and TM modes could be modulated by the plasmonic medium when a plasmonic pad was introduced in the p-type layer, either through reflection or an outcoupling effect [Fig. 4(d)] [35,36]. The excitation of surface plasmon polaritons (SPPs) in metal films typically requires strict conditions, such as matching the frequency and momentum of photons [37]. However, in this work, the edge region of the plasmonic pads should play a crucial role in excitation of plasmonic resonance. Photons, particularly those in TM mode, can effectively excite surface plasmon polaritons as well as local surface plasmon (LSP) resonances at the edge boundaries of plasmonic pads due to the structural features. These resonances assist in altering the wavevector of photons and promoting light extraction through the plasmon outcoupling process. This manipulation mechanism is also consistent with the near-field intensity distribution results in Fig. 3, showing that the most significant enhancement in surface emission occurs at the edge position of the metallic pads. In addition, the deposited Al film with intrinsic nanoscale roughness ($\mathrm{RMS}\sim 2\mathrm{nm}$) should also contribute to the excitation of plasmon resonance, particularly the local SPPs [38], and is conducive to the decoupling of plasmons to enhance light extraction. The corresponding simulations in Figs. 4(e) and 4(f) well demonstrate that the emitted light from the dipole source could be enhanced after the manipulation by plasmonic pads. Taking the emission mainly occurring in the surrounding area of the pads into consideration, the scenario of a dipole source located at the edge region was further evaluated. It can be observed from the figure that for a typical TM mode dipole source when located at a certain distance (0, 2, and 3 μm) from the edge region of the pad, its photons’ transmission still could be modulated and enhanced. This explains the phenomenon of no hole injection and light emission in the etched p-GaN region but the plasmonic pad still could modulate the photons’ propagation. It is because a large number of photons propagate within the LED before being emitted or absorbed; the plasmonic pads should have a much larger spatial manipulation range near the pads for the photons.

The omni-directional reflective properties for the plasmonic pads were also verified by the angle-dependent reflective spectra when using a plane wave source in the FDTD simulation, as illustrated in Fig. 5(a). It was found that the calculated reflectance from the plasmonic pad shows a full-angle enhancement [Fig. 5(b)]. This is consistent with the results in Fig. 2(d) where omni-directional emission enhancement was realized in the Plasmonic LEDs. The extracted near-field distribution for the typical incident light at 0° and $-30°$ from the FDTD simulation well demonstrated the omni-directional reflective performance for the plasmonic pads [Figs. 5(c) and 5(d)].

The enhanced light extraction on TE and TM modes was also evaluated by the edge-emission of the DUV LEDs using a Glan-Taylor prism-based polarizer to record the TE ($E\perp c$ aix) and TM ($E\parallel c$ axis) EL emission separately, as illustrated in Fig. 6(a). The EL emission spectra for the Control sample in Fig. 6(b) showed that the collected emission was dominated by TE mode and the TM mode was relatively weaker. Considering that the ratio of TE mode emission to TM mode emission of sub-250 nm LEDs should be close or TM mode emission is stronger [39], this result indicated that a large portion of TM mode emission in the device fails to escape from the device. For the Plasmonic sample, both the TE and TM mode emissions were enhanced when compared with the Control one, showing a 16.6% and 29.4% increment, respectively [Fig. 6(c)]. The result was consistent with the discussion in Fig. 4 that the photons’ transmission could be manipulated for TE and TM modes by the plasmonic pads towards LEE enhancement. The higher enhanced emission (29.4%) on the TM part revealed that the plasmonic pads have successfully made significant improvements in the TM mode photons’ extraction, which had difficulty escaping in the sub-250 nm DUV LED. The degree of polarization (DOP) was calculated using the equation of $\rho =(I_{\mathrm{TE}}-I_{\mathrm{TM}})/(I_{\mathrm{TE}}+I_{\mathrm{TM}})$ [39], where $I_{\mathrm{TE}}$ and $I_{\mathrm{TM}}$ are the integrated EL intensity for the collected TE and TM mode edge-emission, respectively. The results are summarized in Fig. 6(d) and indicate that a decrease in polarization (from 5.8% to 0.6%) was realized after introducing the plasmonic structure in the devices.

To further enhance the light extraction, a quartz lens bonded by fluorine resin on the devices was introduced to resolve the problem of large internal reflection loss of photons from the sapphire substrate side for the flip-chip LEDs. Thanks to the high DUV light transmittance and good bonding performance of fluorine resin, the quartz lens can be bonded to the chips to adjust the refractive index condition of the light output interface. Even though the refractive index of the fluorine resin is 1.34, the interface layer after bonding can still achieve the grading from the higher refractive index of sapphire (1.82) to the lower refractive index of the quartz lens (1.50), thereby increasing the critical angle of total reflection and significantly reducing the light loss of the chip caused by total internal reflection. The results show that after the introduction of the quartz lens bonded by fluorine resin, the optical output power of the chip has been greatly improved ($\sim 36\%$), and the optical output power of the typical device has been increased from 2.55 mW to 3.45 mW when the device was driven at a current of 100 mA [Figs. 7(a) and 7(b)]. Given the relative low work voltage of 8.70 V, a high WPE of about 0.40% has been realized on the 249.5 nm DUV LED devices.

In summary, to address the light extraction issue of sub-250 nm DUV LEDs, a distributed p/n interdigitated contact flip-chip device structure was designed and embedded with an omni-directional reflective plasmonic pad array in the p-GaN layer to enhance the LEE. Device simulation and characterization showed that the introduction of the plasmonic microstructure nearly did not affect the device’s current injection and $I\text{-}V$ characteristics. However, it significantly improved the photons’ extraction around the plasmonic pads both for TE and TM modes, thereby increasing its overall light output power. The DUV LED achieved the maximum light output power enhancement of 12.5% compared to the Control sample when using an optimized size of plasmonic pads. Further near-field EL emission patterns and FDTD simulation of the device indicated that the light extraction enhancement mainly originated from the plasmonic pads’ modification over the wavevector of the emitted photons from the surrounding quantum wells, thus leading to the increase in the photons’ escape towards the substrate side. In addition, with the introduction of a fluorine-resin-bonded lens, the DUV LED achieved a higher efficiency performance with an output optical power of 3.45 mW and a center wavelength of 249.5 nm driven at 100 mA current. This work demonstrated that specific plasmonic microstructures could effectively manipulate the photon propagation modes in high-Al-composition AlGaN LEDs, thereby enhancing overall light output power, which is crucial for the development of efficient short-wavelength DUV LEDs and addressing key challenges in light extraction. Further optimization of the plasmonic structures’ distribution, density, and passivating materials within the p-type contact layer also holds the potential to boost device performance even more, warranting additional investigation.