Since ultraviolet (UV) light sources have demonstrated wide applications in sterilization and disinfection, phototherapy for skin diseases, secondary plant metabolite promotion, curing, and so on, the portable and nontoxic AlGaN-based UV light-emitting diodes (LEDs) have garnered significant interest in recent years. (1−14) UV light is typically divided into three regions depending on wavelength: UVA (315–400 nm), UVB (280–315 nm), and UVC (<280 nm), with each region serving specific applications. (15) For instance, UVC light is predominantly used for sterilization and disinfection due to its ability to directly damage the protein structures and destroy DNA to hinder the transcription and replication processes. (16−18) Meanwhile, UVB light is valuable for water purification and phototherapy for skin diseases, (19−21) whereas UVA light has found wide applications in various curing processes. (22)

The appropriate combination of UV light with different wavelengths can achieve multifunctional integration to meet the requirements of multiple and complex application scenarios, especially when the emission intensity distribution at different wavelengths can be controlled. In addition, reports suggest that combining UVC with UVB or UVA light has potential synergistic effects on sterilization compared with UVC light alone, offering the promise of enhanced sterilization results. This promising phenomenon was realized not only in water disinfection to sterilize pathogenic microorganisms (19) but also in food decontamination to sterilize foodborne pathogenic bacteria. (23−25) However, the availability of monolithic single-chip multiwavelength UV-LEDs is limited. Most UV light sources with wavelengths spanning over multiple regions are obtained by assembling multiple LED chips with different wavelengths, for instance, water purification process uses multiple types of discrete UV-LED chips simultaneously. (26−28) Nevertheless, these assembled multichip UV light sources not only require different high-cost MOCVD growth runs but also need individual control circuit modules at the expense of device robustness and compactness. Therefore, achieving multiwavelength emitting UV light on a monolithic chip is important to realize low-cost and compact multifunctional UV-LEDs. Recently, it has been shown that the phase separation in AlGaN alloys has the potential to achieve dual-wavelength UV light. (29−32) In AlGaN alloys, Ga adatoms tend to incorporate into the sidewalls of AlGaN terraces, (33) which gives rise to prominent bandgap fluctuations in multiple quantum wells (MQWs), especially when they are grown on large off-cut substrates and thereby realize the dual-wavelength emission. Although the phase separation in AlGaN alloys can be used to generate dual-wavelength UV emission, the growth controllability, emission wavelength, and spectrum tunability are very poor, depriving them from practical applications. To achieve a precise modulation for each peak, it is necessary to grow MQWs with different target wavelengths, similar to the approach used in III-nitride based dual-wavelength visible LEDs. (34−37) However, unlike visible LEDs, the challenges in AlGaN-based UV-LEDs are compounded by the low hole concentration and poor hole mobility in AlGaN materials with high Al content. These limitations hinder the effective transport of holes to the target MQWs region and lead to difficulty in realizing a dual-wavelength UV-LED. Furthermore, whereas most dual-wavelength visible LEDs exhibit variable spectral profiles as the current increases to cover a wide color gamut, such variability is less advantageous for UV-LEDs. The main reason for this is that UV-LEDs often require specific and stable wavelengths for their intended applications. As a result of those challenges, there have been rare reports of monolithic multiwavelength UV-LEDs to date.

Recently, substantial progress has been devoted to improving the efficiency of UVC- and UVB-LEDs. Efficiency improvements were achieved through optimizing the AlN template, (38) multiple quantum well (MQW) structure, (39) p-AlGaN hole injection layer, (12) and so on. In addition to epitaxial growth and structural optimization, research on chip fabrication process optimization has delved into advanced techniques, such as the utilization of finger-like electrodes, (40) high-reflectivity electrodes, (41) flip-type chips, (42) and so on.

In this work, we employed a high-quality AlN template (43) and introduced an Al_0.85Ga_0.15N/Al_0.75Ga_0.25N multiple quantum barrier (MQB) electron blocking layer (EBL) to enhance the “effective” barrier height of the EBL, thus curbing the electron overflow. During device fabrication, we utilized finger-like p-electrodes (Ni/Au (50/100 Å)) to promote the current spreading and a flip-type chip configuration to enhance the light extraction efficiency of UV-LED chips.

The LED structure was grown on an AlN template prepared on a 2 in. (0001)-oriented nanopatterned sapphire substrate (NPSS). The AlN template growth process started with a 30 nm-thick AlN nucleation layer deposited by physical vapor deposition (PVD). Subsequently, a 200 nm-thick 3D-mode AlN layer was grown at medium temperature (1020 °C) followed with a 2 μm-thick epitaxial lateral overgrowth (ELOG) of the AlN layer grown at high temperature (1250 °C). Finally, an 800 nm-thick 2D-AlN layer was grown at 1200 °C.

The growth process of UV-LEDs began with the deposition of a 180 nm-thick AlN homoepitaxial layer on the AlN template. This step was carried out at a temperature of 1250 °C under 35 Torr. Following that, a 220 nm-thick AlGaN transition layer, with an average Al component of 80%, was grown at 1150 °C under 40 Torr. Subsequently, a 1.5 μm-thick Si-doped Al_0.6Ga_0.4N layer was grown at 1130 °C under 40 Torr. The active region was then grown at 1100 °C under 40 Torr. Afterward, three periods of Al_0.85Ga_0.15N(1.5 nm)/Al_0.75Ga_0.25N(2 nm) MQB-EBL and a 40 nm-thick p-Al_0.6Ga_0.4N layer were grown at 1120 °C under 40 Torr. Finally, an 8 nm-thick p-GaN layer was grown at 960 °C under 200 Torr.

The active region consists of both UVB quantum wells (QWs) and UVC QWs for dual-wavelength emission. Regarding the different arrangement of QWs, two types of active region structures were considered, as shown in Figure 1a. In general, for bottom-emitting UV-LEDs, a UVB-upper design, with the UVC quantum well placed beneath the UVB quantum well (i.e., close to the n-type layers), is preferred to minimize internal light absorption (the right graph in Figure 1a). However, as is known, because of the significantly low concentration and weak mobility of holes, (44,45) the electroluminescence in III-nitride LED is mainly contributed from the quantum well closing to the p-type layer, (46) and this phenomenon becomes more pronounced in higher Al-content AlGaN. (47) Besides, this UVB-upper configuration (the right graph in Figure 1a) introduces a high potential barrier (ΔE₂ in Figure 1a) for the transport of holes to UVC QWs. Both of the above reasons may cause the UVB-upper structure to fail to produce controllable and efficient UVC emission and thus dual-wavelength emissions.

To promote the spreading of holes in the MQWs and thereby achieve dual-wavelength emission in UV-LEDs, we propose a UVC-upper active structure (the left graph in Figure 1a) where the UVC quantum well is placed close to the p-type region, aiming to reduce the hole transport barrier (ΔE₁ < ΔE₂) (Figure 1a). As a demonstration, the energy band diagrams calculated at thermal equilibrium of UV-LEDs with UVB- and UVC-upper MQWs configurations (as illustrated in Figure S1 of the Supporting Information) exhibit the difference in the hole transport barrier.

As a result, the active region of our structure consists of threefold UVB MQWs (10 nm-thick Al_0.7Ga_0.3N quantum barrier and 1.6 nm-thick AlGaN quantum well) and a UVC SQW (1 nm-thick ultrathin AlGaN quantum well), as shown in Figure 1b. The Al content of the whole AlGaN quantum barrier is increased to improve the carrier confinement capability of the MQWs, which helps to effectively balance the carrier distribution under a high current injection and thus improves the spectral profile stability. The barrier between the UVB MQWs and the UVC SQW is set as thin as 4.5 nm to promote the transport of holes from the UVC SQW to the UVB MQWs. The above wafer was processed into 20 × 20 mil² (1 mil = 25.4 μm) flip-type chips, named DW-LED. The flip-type chip fabrication process includes mesa etching, n-metal contact deposition and annealing, top p-metal contact deposition and annealing, SiO₂ dielectric layer deposition, pad deposition, chip thinning, and dicing. Photolithography, inductively coupled plasma reactive ion etching (ICP-RIE), plasma enhanced chemical vapor deposition (PECVD), electron beam evaporation, and rapid thermal annealing were employed during these processes. The fabricated wafer was diced into 20 × 20 mil² chips and then mounted on an AlN metalized ceramic substrate for chip characterization using AuSn solder.

First, an experimental comparison is carried out between a UVC-upper-based device (the left graph in Figure 1a) and a UVB-upper-based device (the right graph in Figure 1a). The EL spectra for the bare chips were obtained in the bottom-emission configuration, namely, collecting the EL spectra from the light extraction side using a fiber coupled miniature spectrometer (STS-UV, OceanOptics). As shown in Figure 2, the UVB-upper-based device only exhibits a UVB emission at about 307 nm under a current of 100 mA (75 A/cm²). This is attributed to the high hole potential barrier height that limits the injection of holes into the UVC MQWs, as described in Section 2. In contrast, with the UVC-upper MQW structure, both DW-LED and REF-LED exhibit dual-wavelength emission with UVC emission peaked at around 278 nm and UVB emission peaked at around 307 nm under 100 mA. Although the UVB QWs are located at the far end of the hole injection layers, their UVB luminescence components take a considerable part in the emission spectra, which indicates that the UVC-upper MQW structure effectively facilitates the hole injection into the UVB MQWs. By employing this UVC-upper MQW configuration, the dual-wavelength UV-LEDs offer the flexibility to adjust the wavelength distribution. This can be accomplished by replacing the UVB MQWs with MQWs targeting other peaks, such as UVC, UVB, or even UVA. Note that the EL spectrum of DW-LED exhibits a higher UVC component compared to that of REF-LED. This should be attributed to the utilization of the ultrathin UVC SQW of DW-LED, which will be discussed later.

The current-dependent light output power (LOP), voltage, and WPE of the 20 × 20 mil² chips for DW-LED and REF-LED are measured by a photoelectric measurement system equipped with a 30 cm-diameter integrating sphere (ATA-500, Everfine). As shown in Figure 3a, with increasing injection current from 10 to 100 mA, the LOP increases steadily from 2.04 to 18.60 mW for DW-LED and from 2.02 to 16.92 mW for REF-LED, respectively. Meanwhile, as shown in Figure 3b, the operating voltage exhibits an increase from 5.28 to 6.06 V for DW-LED and from 5.41 to 6.17 V for REF-LED, respectively. The corresponding WPE (WPE = $\frac{\mathrm{LOP}}{I\times V}$) decreases from 3.86 to 3.07% for DW-LED and from 3.73 to 2.74% for REF-LED, respectively, as shown in Figure 3c. The decreasing trend of WPE with increasing injection current is common in DUV-LEDs, which may be related to electron leakage under high current injection. (48,49) It is worth mentioning that both DW-LED and REF-LED exhibit high maximum WPE values of 3.86 and 3.73% under 10 mA, respectively, which are comparable to the values reported for single-peak UVB-LEDs and UVC-LEDs. (43,50,51)

Subsequently, the current-dependent spectral profile stability of dual-wavelength UV-LEDs is studied. The EL spectra under variable current injection (ranging from 10 to 100 mA) for DW-LED and REF-LED are obtained in the bottom-emission configuration, and the results are shown in Figure 4a,c, respectively. The UVB and UVC components in each emission spectrum of DW-LED and REF-LED are extracted and plotted versus current in Figure 4b,d, respectively. It is noted that this specific range of current injection is well-suited for high-efficiency UV-LED operation, as WPE typically decreases with increasing current for UV-LEDs. As shown in Figure 4a,b, as the injection current increases, the integrated intensity of both UVB and UVC emissions in DW-LED shows a nearly linear increase, indicating a relatively uniform carrier distribution and recombination between these two types of quantum wells. On the contrary, for REF-LED, as the injection current increases, its integrated intensity of UVC emission quickly saturates (Figure 4c,d), whereas its UVB component keeps a similar linear increasing tendency to the UVB component of DW-LED, as shown in Figure 4b. As a result, as shown in Figure 4e, with increasing injection current from 10 to 100 mA, the spectral profile of REF-LED shows a large variation, with the UVC emission ratio (UVC/(UVB + UVC)) decreasing from 38 to 17%. This exhibits an apparently more unstable current-dependent feature than DW-LED, whose UVC part only slightly decreases from 53 to 48%.

The significantly more stable spectral profile of DW-LED with increasing current can be attributed to the use of an ultrathin UVC SQW. In Al-rich AlGaN quantum wells, a strong built-in electric field causes the electron and hole wavefunction to become spatially separated, resulting in a greatly reduced radiative recombination rate. (52,53) The use of an ultrathin UVC SQW in DW-LED allows for a larger electron–hole wavefunction overlap, leading to a higher radiative recombination rate and thereby a stronger UVC emission component and a higher WPE of the LED chip compared to REF-LED. (54,55) We also measured the photoluminescence (PL) spectra of dual-wavelength MQWs of DW-LED and REF-LED at both 10 and 300 K using a 213 nm excitation laser to estimate the internal quantum efficiency (IQE), as shown in Figure S4 of the Supporting Information. Our results indicated similar IQEs of UVB emissions (45.1% for DW-LED and 51.3% for REF-LED) and significantly different IQEs of UVC emissions (52.2% for DW-LED and 21.1% for REF-LED). This discrepancy in IQE can be largely attributed to the utilization of the ultrathin UVC SQW in DW-LED, which contributes to a higher radiative recombination rate and thus an enhanced IQE. Considering that the UVC SQW of REF-LED exhibits a low radiative recombination rate and low IQE, a large fraction of holes cannot be captured and thus migrate into the UVB MQWs to realize radiative recombination. Therefore, this may contribute to a smaller discrepancy in WPE than the tested IQE for DW-LED and REF-LED. The superiority of utilizing ultrathin QWs is also evidenced by the energy band diagram simulation in Figure S5 of the Supporting Information, where a larger electron–hole wavefunction overlap is demonstrated in the ultrathin QWs compared to that in the conventional QWs. On the other hand, owing to the lower radiative recombination rate, the emission intensity of the conventional UVC SQW in REF-LED quickly saturates under a small current injection, in which case the UVB emission intensity shows a sustained increase with increasing injection current (as shown in Figure 4c,d), inducing an unstable spectral profile. In contrast, the larger radiative recombination rate of the ultrathin UVC SQW in DW-LED results in a much slower saturation trend in the emission intensity with increasing current injection, as shown in Figure 4a,b. This consequently leads to a more stable current-dependent spectral profile.

After obtaining a dual-wavelength UV-LED with stable current-dependent spectral profile, we propose a method to achieve tunability of the emission distribution by adjusting the thickness of the potential barrier between the UVB MQWs and the ultrathin UVC SQW (referred to as a modulation layer). Through adjustment of the thickness of the modulation layer, the hole injection into the UVB MQWs can be modified, which has a significant effect on the spectral profile. To demonstrate and verify this method, we fabricated two other samples (called DW-LED2 and DW-LED3, herein) with modulation layer thicknesses of 3.5 and 6.5 nm, respectively, in addition to sample DW-LED, which has a modulation layer thickness of 4.5 nm. Figure 5a shows their EL spectra under 100 mA current injection. It is evident that samples DW-LED, DW-LED2, and DW-LED3 have different spectral profiles regarding their relative UVB/UVC emission intensity. To facilitate a direct comparison, we normalized the EL spectra to their UVC peak emission intensity, as shown in Figure 5b. Our results show that, by decreasing the modulation layer thickness from 6.5 to 4.5 and 3.5 nm, the proportion of UVC emission decreases from 72 to 48 and then to 30%. This wide range of adjustability is attributed to the fact that, as the potential barrier becomes thinner, the hole injection into the UVB MQWs is enhanced, which reduces the proportion of UVC emission. (56) It should be noted that the EL spectra measured in the bottom-emission configuration differ from those obtained through the integrating sphere, as discussed in Figure S6 of the Supporting Information. This discrepancy arises from the fact that the EL spectra obtained in the bottom-emission configuration primarily collect direct radiation from the emission side of the LED chips, where part of the transverse magnetic (TM) polarized light remains undetected. Nevertheless, it is worth highlighting that the superiority of our design for dual-wavelength UV-LEDs is valid in both measurement approaches.

In addition, the current-dependent stability of the EL spectral profiles for DW-LED2 and DW-LED3 is examined. The EL spectra under variable current injection (ranging from 10 to 100 mA) for DW-LED2 and DW-LED3 are obtained in the bottom-emission configuration, and the results are plotted in Figure 6a,c, respectively. The UVB and UVC components in each emission spectrum of DW-LED2 and DW-LED3 are extracted and plotted versus current in Figure 6b,d, respectively. It is observed that both UVB and UVC emission intensities for DW-LED2 and DW-LED3 show a steady increase with injection current. The UVC emission ratio remains around 72% for DW-LED3 and slightly decreases from 38 to 30% for DW-LED2 with increasing injection current from 10 to 100 mA. As the modulation layer thickness decreases from 6.5 to 4.5 and 3.5 nm, the UVC emission ratio shows a more prominent decreasing trend with increasing injection current, from around 0 to 5 and then to 8%. This trend is reasonable because the hole confinement of the UVC SQW weakens as the potential barrier thins, which enhances the hole injection into the UVB MQWs and reduces the saturation emission intensity of the UVC SQW. The maximum WPEs show values of 4.20 and 3.43% for DW-LED2 and DW-LED3, respectively (see Figure S7 of the Supporting Information). A slight increase in WPE is observed as the modulation layer thickness decreases, that is, the maximum WPE increases from 3.43 to 3.86 and then to 4.20% as the modulation layer thickness decreases from 6.5 to 4.5 and then to 3.5 nm. This can be related to a reduction in the thickness of the quantum barrier. The thinning of the quantum barrier mainly has two effects: On the one hand, it reduces the QCSE, leading to increased electron–hole wavefunction overlap and enhanced radiative recombination rate. (57) This could contribute to an improved WPE level. On the other hand, it reduces the potential barrier to hole migration. This, in turn, enhances hole spreading, which is also helpful in promoting WPE. (58,59) As a result, WPE is improved in the LEDs with thinning of the modulation layer. Overall, similar to sample DW-LED, both samples DW-LED2 and DW-LED3 show a high WPE value and outstanding current-dependent stability, which fully demonstrate the possibility of achieving high-performance dual-wavelength UV-LEDs with a stable and controllable spectral profile. Furthermore, it is worth noting that, in this work, we focused on the MQW design to achieve stable and controllable dual-wavelength emission. Although our current results are promising, there is room for further improvement in performance. This improvement aligns well with the ongoing advancements in UV-LED technology, such as improving the hole injection capability of high Al-content p-type AlGaN, increasing light extraction efficiency, and exploring other potential avenues.

Subsequently, we compare our achievement with those of previously reported dual-wavelength LEDs. Table 1 shows that previous reports have mainly focused on dual-wavelength LEDs in the visible light spectrum, aiming to achieve variable spectral profiles with varying currents to cover a broad color gamut. In our work, the stacking order of UVC and UVB QWs exhibits a close dependency on the spectra distribution, especially the number of emission peaks, whereas the stacking order of the MQWs of the visible LEDs has relatively less impacts. Our achievement is based on the design of MQW stacks and highlights a remarkable feature: the ability to achieve a stable and controllable emission distribution within a specific range of current injection. This feature holds importance because the applications of UV light rely on specific and stable wavelengths.

Finally, to investigate the UVB MQW absorption effect on UVC light, the current-injection emission spectrum is measured from both the front and back sides of the UVC-upper LED wafer using an on-wafer indium-dot quick test. As shown in Figure 7, the relative intensity between UVB and UVC emissions from the back side does not show a significant change compared to that from the front surface under a specific 100 mA current injection. This indicates that the UVC emission from the back side is not significantly reduced compared to the UVB emission. Therefore, the aforementioned light absorption in the UVC-upper design can be considered negligible in our practical device.