1. Introduction

AlGaN-based ultraviolet-C light-emitting diodes (UV-C LEDs) have attracted increasing attention owing to their numerous advantages, including environmental friendliness, energy efficiency, high power density, and long operational lifetime, in comparison to mercury-based lamps. Notably, UV-C LEDs emitting wavelengths below 280 nm possess the ability to directly induce damage to the DNA or RNA of bacteria and viruses. Consequently, they hold significant potential for applications in water sterilization, air purification, and related domains [1–3]. Nonetheless, a primary limiting factor currently impacting UV-C LEDs is their exceedingly low external quantum efficiency (EQE), typically not surpassing 10%. This deficiency in EQE arises from inadequate internal quantum efficiency (IQE) and light extraction efficiency (LEE) [4–6].

The LEE of Al-rich UV-C LEDs is significantly influenced by the unusual sub-valence-band ordering observed in the Al-rich AlGaN/AlGaN MQW active regions. In the case of [0001]-oriented Al-rich MQWs, the generated UV-C photons exhibit strong polarization, where transverse magnetic (TM) (E//c) polarized photons demonstrating lateral propagation, while transverse electric (TE) (E⊥c) demonstrating vertical propagation [7,8]. Additionally, the presence of an absorptive GaN layer in UV-C LEDs contributes to the low LEE for wavelengths shorter than 280 nm. The limited availability of low-absorption materials for UV-C photons poses an additional challenge to the LEE process. As a result, commercial UV-C LEDs employ flip-chip structures with reflective mirrors and utilize nanostructured p-contact to mitigate light absorption [9,10]. Various well-established techniques, including sapphire roughening, nano-patterned substrates, Al reflectors with distributed Bragg reflectors (DBR), and thermally oxidized sidewalls, have been developed to enhance LEE [11–16]. A novel approach involves the direct deposition of MQWs on an AlN template in transverse-structure LEDs, resulting in a double increase in luminous efficiency compared to the conventional LED structure with an n-AlGaN layer [17]. However, these methods are not effective for AlGaN-based UV-C LEDs due to the parallel propagation of TM-polarized UV-C photons with the substrate [18–20]. The experimental investigation of the TE and TM modes in AlGaN-based UV-C LEDs necessitates extensive spatial decomposition. Previous studies have indicated that roughening the sapphire sidewall enhances the extraction efficiency of UV-C LEDs, applicable to both TE and TM mode light without polarization selectivity [21]. The effectiveness of a novel moth-eye microstructure fabricated on the back of a sapphire substrate in improving the LEE of UV-C LEDs has been demonstrated [22]. Moreover, the challenges arising from the presence of numerous threading dislocations and the extremely low LEE of TM-polarized emission can be addressed through the implementation of a flip-chip structure with AlGaN quantum-well heterostructures grown on a hexagonal nano-patterned sapphire substrate (NPSS) and an inclined Al mirror [23]. Angle-resolved cathodoluminescence (ARCL) proves to be a powerful tool for collecting the angular distribution of emission and determining the polarization of nanostructured UV-C LEDs [24,25]. Furthermore, there have been reports suggesting that an appropriate inclined MESA angle can significantly facilitate the extraction of photons into the air [26]. Research findings have demonstrated that an angle of 37.83° is an optimal inclination, resulting in a 48% enhancement in the optical power of UV-C LEDs [27]. In AlGaN QWs, emitted light occurs in either the TE or TM mode. With an increase in the Al composition, the proportion of TM-polarized light increases [28]. However, the spatial distribution of TM and TE modes within the QWs remains unclear.

Our primary focus in this study is to optimize the inclination angle (θ) of AlGaN MQW layers to enhance the LEE for UV-C LEDs. To achieve our goal, we used the focused ion beam and scanning electron microscopy (FIB-SEM) to fabricate UV-C epitaxial samples with various θ, the polarization characteristics of which are analyzed by ARCL technique. Combining numerical simulations and experimental analyses, we explored the sidewall light extraction mechanism and successfully distinguished the contributions of TM and TE photons to the overall luminescence of the UV-C MQW structure.

2. Experiments

Figure 1(a) presents a detailed schematic of the experimental setup used in this study. The electrons accelerated in a TESCAN SOLARIS (S900) FIB-SEM system, are utilized to excite the CL signal of sample through a 600-um diameter aperture of a parabolic mirror. This parabolic mirror collects the emitted CL of sample positioned directly above its focal point and directs it out of the vacuum chamber through a glass vacuum flange. The resulting beam from the paraboloid is then directed onto a Peltier-cooled (−70°C) back-illuminated silicon charge-coupled device (CCD) camera with a resolution of 1024 × 1024 pixels. In ARCL spectroscopy, the CL emission is collected from the entire beam incident on the parabolic mirror. Subsequently, this emitted light is redirected out of the vacuum chamber toward a 2D CCD imaging array, where each pixel within the acquired image corresponds to a unique emission angle. These emission angles are defined by a zenithal angle (α) ranging from 0° to 90° and an azimuthal angle (φ) ranging from 0° to 360°. Every point in the beam profile represents a specific combination of these angles. The resulting image allows for the direct determination of the angular distribution of CL emission, facilitating optical imaging and spectroscopy at a deep subwavelength spatial scale [24].

 

Fig. 1. (a) Schematic diagram of the experimental setup. (b)The TEM image shows the UV-C epitaxial layer, with highlighted red lines indicating the distinct layers: the P layer, the MQW layer, and the N layer. (c) Cross-sectional view of the UV-C epitaxial layers with θ. Blue arrows show the propagation directions of TE- and TM-polarized photons.

Download Full Size | PDF

An AlGaN-based UV-C epitaxial wafer was grown on a (0001) sapphire substrate using a low-pressure metal-organic chemical vapor deposition (MOCVD) system. The structure of the UV-C LED includes an AlN buffer layer, an undoped Al_0.45Ga_0.55N layer, and an n-Al_0.6Ga_0.4N contact layer doped with Si. The active region comprises five pairs of Al_0.45Ga_0.55N (5 nm)/Al_0.57Ga_0.43N (15 nm) MQWs. An electron blocking layer (EBL) of Mg-doped p-Al_0.8Ga_0.2N with a thickness of 25 nm was deposited between the MQW layer and Mg-doped P-GaN layer with a thickness of 40 nm, as shown in Fig. 1(b) by the image captured through transmission electron microscope (TEM). The N-AlGaN layer has a total thickness of 2 µm, and Fig. 1(b) only depicts a partial section of it. Figure 1(c) presents a schematic cross-sectional view of the UV-C epitaxial wafer, where the θ is precisely etched at six angles (15°, 25°, 35°, 50°, 65°, and 80°) using the FIB, with each pit having an etched depth and width of approximate 5µm (D) and 10 µm (L₁), respectively. L₂ represents the inclined sidewall, which varies with θ. Based on previous research [27], we opted to etch these six values of inclined angles. When measuring the CL signal, the sample is mounted with the inclined surface facing upward. Additionally, due to the existence of TE and TM polarization modes in the emitted photons from the quantum well, TE-mode photons primarily propagate along the c-axis. Only a limited number of photons reach the sidewall, with a portion being absorbed by the top P-GaN layer. In contrast, TM-mode UV-C photons mainly propagate vertically (perpendicular to the c-axis), allowing most photons to effectively reach the sidewall of the device. It is known that for UV-C LEDs, radiative recombination in the MQWs is primarily dominated by the C-CH interband transition, resulting in the generation of TM-polarized photons [7]. The blue arrows represent the propagation directions of TM-polarized photons and TE-polarized photons. Therefore, it is necessary to optimize the sidewall θ to maximize the effective extraction efficiency of TM-mode photons, thereby achieving the maximum sidewall LEE.

3. Results and discussion

Figure 2 illustrates the CL spectra of the MQW layers excited by 10 KV electrons beam, showcasing different θ. It can be observed that the highest CL intensity is achieved at a θ of 35°, while the CL intensity decreases as the angle exceeds 35° or is below 35°. Hence, it becomes clear that the sidewall's θ in UV-C epitaxial layer significantly affects its emission intensity. Consequently, extensive experimentation is crucial to pinpoint the optimal θ for maximizing emission intensity. The forthcoming discussion will delve into elucidating the underlying mechanism that gives rise to the variations in optical power.

 

Fig. 2. CL spectral intensity distribution under different θ.

Download Full Size | PDF

In Fig. 3(a), intensity variation curves of α within the range of 0° to 90° with 15° intervals are depicted for three distinct inclined angle θ: 25°, 35°, and 80°. It is evident from the plots that at θ = 35°, the intensity is consistently higher compared to the other two θ. Furthermore, at θ = 35°, the emission is relatively uniform with stronger intensity for each φ angle in the α range of 0° to 45°, as indicated by the purple circle. Notably, a pronounced dip is observed in the α range of 45° to 90°. Additionally, the intensities of CL exhibit symmetry at certain ranges of α, such as α < 45°, while particularly strong intensities are observed at φ = 120° and 240°. In Fig. 3(b), intensity variation curves of φ within the range of 70° to 250° with 30°intervals are presented for the same three θ. At θ = 35°, the emitted intensity is higher than the other two θ, especially exhibiting a stronger intensity in the range of α from 0° to 15°, a peak around α ≈ 40°, and another elevated intensity in the range of α from 80° to 90°.

 

Fig. 3. (a) Experimental data presenting the CL intensity at different θ, covering the range of α from 0° to 90° with a 15° interval. The data are plotted as a function of φ. (b) Experimental data presenting the CL intensity at different θ, covering the range of φ from 70° to 250° with a 30° interval. The data are presented as a function of α.

Download Full Size | PDF

In order to further elucidate the underlying mechanism behind the ARCL, we conducted numerical simulations using the finite difference time domain (FDTD) to investigate the relative proportions of TM-polarized and TE-polarized photons in the MQW layers at different θ. As shown in Fig. 4(a), the model is constructed based on the structure depicted in Fig. 1(c). The positions of TM and TE mode sources were centered within the quantum well layer. The important parameters used in the FDTD simulations, such as refractive index, can be found in Refs. [29] and [30]. To conveniently compare with the experimental data, the model rotates 90⁰ so that the inclined sidewall faces upward. During the simulation process, we arranged a matrix with six rows, each comprising 36 monitors, covering the entire model to collect the emitted photons. The xyz coordinate system indicates the ranges of α and φ angles, which correspond to the ranges of 0°-90° in the zy-plane (α) and 0°-360° in the xy-plane (φ). The range of α from 0°-90° is divided into six sections, each section separated by 15 degrees. Figures 4(b) and (c) display TM and TE photons in xy plane with all photons accumulated along α angle. Each monitor separates the xy-plane by 10 degrees. From the simulated data in Fig. 4(b), when only TM sources are present, it can be observed that at φ = 60°, 120°, 240°, and 290°, corresponding to the sidewalls of the model, the intensity of TM-polarized photons at these positions is relatively higher compared to TM photons at other φ. Figure 4(c) presents the simulation results for TE-polarized photons, revealing a complementary trend compared to TM-polarized photons, i.e., lower intensity at positions on the sidewalls (60°, 120°, 240°, and 290°), whereas higher intensity at positions on the surface side (20°, 160°, 220°, and 340°). This phenomenon arises because for UV-C LEDs, radiative recombination in the MQWs is primarily driven by the C-CH interband transition, which generates TM-polarized photons [7].

 

Fig. 4. (a) The front view of the FDTD simulation model, along with a xyz coordinate system indicating the range of α and φ. The FDTD simulation results for (b) TM-polarized photons and (c) TE-polarized photons in the xy-plane, with all photons accumulated along α angle.

Download Full Size | PDF

Figure 5(a) presents the ARCL intensity as a function of φ for different θ in the experiment. These findings further corroborate the observations described in Fig. 2, where a θ of 35° yielded the highest emission intensity. However, at θ of 50°, 25°, 15°, and 65°, and 80°, the CL intensity gradually decreased, primarily due to limited extraction of TM-polarized photons. In Fig. 5(a), the polarization is evident at most inclined angles of θ. Additionally, particularly strong intensities are observed at φ = 120° and 240°. Therefore, we employ FDTD simulation to meticulously differentiate the contributions of TM- and TE-polarized photons. The trends are in accordance with the FDTD simulation as illustrated in Fig. 5(c), where both TM and TE light sources are activated simultaneously. TM and TE light sources are positioned vertically to each other. We use degree of polarization (DOP) to describe the variation of polarization due to the contribution of TM and TE in the total output. The formula for calculating DOP can be described as [31]:

Here, I_max and I_min represent the maximum and minimum light intensities, respectively. The DOP as a function of θ is depicted in Figs. 5(b) and 5(d), based on experimental data and simulation data, respectively. The trends of both curves are in good agreement. This outcome is attributed to the polarization reduction in the MQW layer at θ = 35°, leading to a more uniform distribution of light in various directions compared to other θ.

 

Fig. 5. The ARCL intensity as a function of φ at different θ for (a) experimental data and (c) simulation data (TM + TE). The calculated DOP for (b) experimental data (d) simulation data (TM + TE).

Download Full Size | PDF

In Fig. 6(a), curves depicting the CL intensity as a function of the monitors along φ, with α ranging from 0° to 90° at 15° intervals, are presented for three specific θ: 25°, 35°, and 80°. Under these three θ, the intensity of α within the range of 0° to 45° is comparatively high and stable for each φ, as indicated by the purple circle. Additionally, fluctuations in intensity and a dip at φ = 180° are observed within the range of α from 45° to 90°, aligning with the trends observed in the experimental data in Fig. 3(a). In Fig. 6(b), curves representing the CL intensity as a function the monitors along α, with φ ranging from 70° to 250° at 30° intervals, are displayed for the same three θ. Each monitor separates the zy-plane (α) by 2.5°. Notably, at θ = 35°, there is a prominent peak in α from 0° to 10°. Another distinct peak is observed around α ≈ 40°, indicated by the blue curve in the range of φ= 160°-190°. An observable increase in intensity is noted in α from 80° to 90°. These simulated observations align with the experimental phenomena, providing validation for further analyses. Therefore, we can comprehensively discuss the contributions of TM and TE by separately plotting their outputs in Fig. 7.

 

Fig. 6. Simulation data (TM + TE) presenting the CL intensity at different θ as a function of the monitors (a) along φ, covering the range of α from 0° to 90° with a 15° interval and (b) along α, covering the range of φ from 70° to 250° with a 30° interval.

Download Full Size | PDF

 

Fig. 7. (a) and (b) represent simulation data with only TM and TE light source as a function of φ, covering the range of α from 0° to 90° with a 15° interval, for θ = 35°. (c) and (d) represent simulation data with only TM and TE light source as a function of α, covering the range of φ from 70° to 250° with a 30° interval, for θ = 35°.

Download Full Size | PDF

Figures 7(a) and 7(b) cover the range of α from 0° to 90° at a 15° interval, plotted as a function of the monitors along φ. Comparing the plot in Figs. 7(a) and 7(b) with that of θ = 35° in Fig. 3(a), it can be concluded that TM photons are the main contributors to the strong emission in the range of α from 0° to 45°, as indicated by the purple circle. This is particularly evident within the α range of 0° to 15°. As depicted in Fig. 6(a) and 7(b), we can confidently conclude that the strong intensities at φ = 120° and 240° observed in Fig. 3(a) are predominantly caused by TM-polarized photons. Figures 7(c) and 7(d) cover the range of φ from 70° to 250° at a 30° interval, plotted as a function of the monitors along α. Again, combining the information from Figs. 7(c) and 7(d) with that in Fig. 3(b), it can be inferred that the peak at α ≈ 10° mostly comes from TM photons, and at α ≈ 40° results from the combined presence of TM and TE photons, as evidenced by the intensity of both at α ≈ 40° in Figs. 7(c) and 7(d). The enhanced emission near α = 90° is attributed to TE photons, as Fig. 7(d) indicates an increasing trend in TE photon emission at α ≈ 80°, especially in the regions of pink φ = 220°-250° and red φ = 100°-130°, providing an explanation for the phenomena observed in experimental Fig. 3(b).

To further differentiate the contributions of TM and TE photons at various angles, Fig. 8 illustrate the ratio of TM-polarized photon intensity to TE-polarized photon intensity in the α = 0°-15° range and the ratio of TE photon intensity to TM photon intensity in the α = 75°-90° range, respectively. Presenting this information numerically enhances the clarity of the proportion of TM and TE photons’ emission intensities at various monitors along φ. The graphs demonstrate that within the α = 0°-15° range (sidewall), TM photon emission surpasses that of TE photons. Conversely, in the α = 75°-90° range (epilayer surface), TE photons dominate.

 

Fig. 8. The ratio of TM-polarized photons to TE-polarized photons in the α range of 0°-15° and TE-polarized photons to TM-polarized photons in the α range of 75°-90°, with overall photons accumulated along φ.

Download Full Size | PDF

4. Conclusion

In this research, we utilized ARCL spectroscopy to explore the optical polarization of (0001)-oriented AlGaN MQW structures in UV-C LEDs. We employed a FIB-SEM system to prepare samples with varying inclined angle θ of sidewall. The investigation revealed that an optimal θ of 35° resulted in the highest CL signal compared to samples with different θ values. Extensive FDTD simulations were conducted to differentiate the spatial distribution of TM- and TE-polarized photons within the MQWs, highlighting specific angles that significantly enhanced the LEE of TM-polarized photons. The simulations also confirm the effectiveness of the ARCL technique in discerning the complementary spatial distribution of TM and TE polarized photons. Near the sidewall angles, CL signal was predominantly dominated by TM photons, which played a crucial role in increasing LEE and reducing the DOP, leading to a more uniform spatial distribution of CL. Remarkably, this approach allows for the analysis of light polarization without the need for a polarizer, facilitating the differentiation between TM and TE modes. This capability provides flexibility in selecting the appropriate emission mode based on specific application requirements. It is suggested that further enhancements in LEE and output power can be achieved by adjusting the spatial distribution angle and geometrical design. Importantly, the presented approach underscores the potential of properly designed inclined sidewalls and strategically placed Al reflectors to develop other deep ultraviolet light emitters with significantly improved performance.