263 nm wavelength UV-C LED on face-to-face annealed sputter-deposited AlN with low screw- and mixed-type dislocation densities

The use of deep-ultraviolet (DUV) light with wavelengths shorter than 300 nm for sterilization and virus inactivation has garnered significant interest. Mercury lamps are widely used as high-power DUV light sources for these applications. However, the demand for alternative light sources remains high because mercury has a critical impact on the environment and the human body. Against these strong market demands, AlGaN-based DUV light-emitting diodes (DUV-LEDs) with emission wavelengths in the UV-C region (200–280 nm) have yielded improvements in external quantum efficiency (EQE), maximum output power, and long-term reliability. ^1–13) Nevertheless, the performance of UV-C LEDs remains significantly inferior to that of InGaN-based LEDs, which have a wavelength within the range of near-UV to blue light. ¹⁴⁾

A problem with AlGaN-based UV-C LEDs is the high density of threading dislocations (TDs) in the AlN and AlGaN grown on sapphire substrates. The TDs act as non-radiative recombination centers, which reduce the internal quantum efficiency (IQE) ¹⁵⁾ and accelerate device degradation. ¹⁶⁾ Thus, high-performance UV-C LEDs mandate the preparation of AlN templates with low TD densities (TDDs). We previously fabricated AlN templates with low TDDs on sapphire substrates combining sputter deposition and post-deposition high-temperature face-to-face annealing. ¹⁷⁾ Hereinafter, this type of AlN template is referred to as face-to-face annealed sputter-deposited AlN (FFA Sp-AlN). The lowest TDD of the FFA Sp-AlN was approximately 4 × 10⁷ cm⁻², ¹⁸⁾ which is one order of magnitude lower than that of typical AlN templates grown on sapphire substrates via metalorganic vapor phase epitaxy (MOVPE). Furthermore, the combination of sputter deposition and high-temperature annealing is a simple and cost-effective process. The lower TDDs of the AlN templates applied to UV-C LEDs on FFA Sp-AlN improved luminous efficiency and extended device lifetimes. ^16,19–22)

However, because the densities of the screw- and mixed-type dislocations in FFA Sp-AlN are extremely low, the spiral growth induced by these screw-component-included dislocations tend to develop into excessively large hillock structures. ²³⁾ These structures create surface roughness on the AlGaN film and deteriorate the optical properties of the UV-C LEDs fabricated on the FFA Sp-AlN. Thus, to exploit the low TDDs of FFA Sp-AlN, it is necessary to establish growth technologies for AlGaN ensuring high surface flatness and minimal lattice relaxation. We previously reported that although hillock sizes can be reduced using sapphire substrates with large off-angles, the luminescence efficiency is lower when the substrate off-angle is excessively large. ²³⁾

In this study, we applied two approaches to suppress hillock formation and realize AlGaN films with high surface flatness, even on substrates with a small off-cut angle of 0.2°, to overcome the drawback of a large substrate off-angle. The first approach is the reduction of the screw- and mixed-type dislocation densities in the FFA Sp-AlN, i.e. hillock density reduction, by controlling the fabrication conditions for the AlN template. The second approach is hillock size reduction by controlling the MOVPE growth conditions for AlGaN. The resulting smooth surface of the AlGaN film improved the luminescence properties of the UV-C LEDs. Moreover, the maximum EQE of UV-C LEDs with a peak wavelength of 263 nm reached approximately 4.9% and 8.0% without and with silicone encapsulation, respectively.

First, we describe the relationship between the sputter-deposition conditions for the AlN films and the density of the screw- and mixed-type dislocations in the FFA Sp-AlN. To fabricate 600 nm thick AlN films, radio-frequency magnetron sputtering was applied to sapphire substrates with an off-angle of 0.2° toward the m-axis of the sapphire. Two types of AlN templates were prepared by applying substrate temperatures of 700 °C or 750 °C. The AlN films were then subjected to 1600 °C–1700 °C face-to-face thermal cycle annealing ²⁴⁾ for 36 h. Additional details regarding the FFA Sp-AlN fabrication conditions can be found elsewhere. ^22,25) Subsequently, a 200 nm thick AlN homoepitaxial layer and 300 nm thick unintentionally doped (UID) Al_0.70Ga_0.30N layer were sequentially grown on the FFA Sp-AlN via MOVPE. The MOVPE growth conditions were identical to those previously reported. ²³⁾ The MOVPE AlGaN growth forms hillock structures from the screw- or mixed-type dislocations in the FFA Sp-AlN. Thus, the hillock structures and screw-component-included dislocations had a one-to-one correspondence. ²³⁾

Figure 1 shows Nomarski microscopy images of the Al_0.70Ga_0.30N samples grown on FFA Sp-AlN under two different sputter-deposition temperature conditions. The hillock density of the sample with higher temperatures was low. At deposition temperatures of 700 °C and 750 °C, the hillock densities were approximately 1.4 × 10⁵ and 4.9 × 10³ cm⁻², respectively. These values are one-to-three orders of magnitude lower than those for the AlGaN grown on conventional FFA Sp-AlN fabricated at a deposition temperature of 600 °C, featuring a hillock density in the order of 10⁶ cm⁻². ²³⁾ We previously reported that the sputter-deposition conditions and resultant crystallinity of an as-sputtered AlN film substantially affect the crystallinity of the post-annealed FFA Sp-AlN. ²⁶⁾ We also elucidated the need for the highly c-axis-oriented growth of sputter-deposited AlN films to minimize the crystalline tilt [X-ray rocking curve full width at half maximum (XRC-FWHM) of symmetric (000n) diffraction] of the FFA Sp-AlN. The occurrence of lower the screw- and mixed-type FFA Sp-AlN dislocation densities at higher deposition temperatures can be attributed to the improvement in the c-axis orientational ordering of the AlN film in the as-sputtered state.

Zoom In Zoom Out Reset image size

Subsequently, Al_0.75Ga_0.25N:Si films with a thickness of 1 μm were grown on FFA Sp-AlN. The FFA Sp-AlN fabrication conditions applied in the experiments were identical to those described previously. ²³⁾ The relationship between the MOVPE growth conditions and the AlGaN:Si surface morphology was evaluated by systematically varying four parameters: growth temperature (T_g), growth rate (R_g), NH₃ partial pressure (P_NH3), and H₂ partial pressure (P_H2). The AlN mole fraction and AlGaN film thickness remained fixed for all experiments to obtain an accurate understanding of the effect of the growth conditions on the surface morphology. However, the AlN mole fraction of AlGaN generally varies according to the growth conditions when the supply amounts of the group-III sources, trimethylgallium (TMGa) and trimethylaluminum (TMAl), are fixed. Furthermore, among the aforementioned four growth conditions, R_g can be affected by the other three growth conditions. For example, an increase in T_g or P_NH3 decreases R_g because it increases the rate of the gas-phase parasitic reaction between TMAl and NH₃. Thus, we adjusted the amounts of TMGa and TMAl supplied in each experiment such that the AlN mole fraction of AlGaN:Si remained constant at 75%, and the other growth conditions remained unchanged when one of the target parameters was altered. The MOVPE growth conditions for Samples A–H are described in Table I. Figures 2(a)–2(h) show the morphologies of Samples A–H, respectively; (a) and (b) show the dependence on T_g, (c) and (d) show the dependence on R_g, and (e)–(g) show the dependence on P_NH3 and P_H2. These results indicate that a high T_g, low R_g, and high P_NH3 can effectively reduce the size of the hillock structure, whereas P_H2 does not significantly influence the surface morphology. As shown in Fig. 2(h), the hillock structures were so small that they were barely detectable on the sample grown under optimized conditions. Notably, the hillock density was independent of the MOVPE growth conditions for AlGaN, as it was determined by the density of the screw- and mixed-type dislocations in the FFA Sp-AlN template.

Zoom In Zoom Out Reset image size

The effect of the growth conditions on the AlGaN film surface morphology can be described as follows: an increase in T_g and a decrease in R_g both reduce the surface supersaturation. The intervals between each step-and-terrace structure formed as a result of spiral growth are known to expand and consequently flatten hillock structures. ²⁷⁾ The influence of P_NH3 can be explained by the surface migration distance of adatoms. For a hillock structure to enlarge, adatoms from the surrounding region must concentrate toward the nucleus of the hillock structures. However, if the surface migration distance is short, only the atoms adsorbed in the vicinity of the hillock structure can reach the hillock structure. In this case, there would be no significant difference between the growth rate of the hillock structure and that of the regions exclusive of the hillock structure, and the dislocation-induced spiral growth would not develop into an excessively large hillock. These conditions promote uniform crystal growth across the entire surface and are believed to have limited the size of the hillocks in this study.

Figures 3(a) and 3(b) show the magnified surface morphology images of 1 μm thick Al_0.75Ga_0.25N:Si grown under the optimum condition [Fig. 2(h)]. The step-and-terrace structures indicate meandering features and some pits at the volleys of the steps. It should be noted that these pits were not correlated with the TDs, which were confirmed via plan-view TEM and comparison of the distributions of the pits and TDs. The step meandering and resultant pit formation are drawbacks of the growth conditions (high T_g, low R_g, and high P_NH3) for suppressing hillock structures. This problem can be overcome by growing an additional 100 nm thick AlGaN:Si interlayer grown with relatively low T_g (1050 °C). Figures 3(c) and 3(d) show the surface morphology of the interlayer grown on the sample shown in Figs. 2(h), 3(a), and 3(b). The AlN molar fraction (75%), R_g (0.4 μm), and P_NH3 (10.0 kPa) of the interlayer are identical to the underlying AlGaN:Si layer. The step meandering and the pits were suppressed on the interlayer. Further, 2 monolayer (ML) steps with ∼5.0 Å height fabricated on the 1150 °C grown AlGaN:Si [Fig. 3(b)] were solved into 1 ML steps with ∼2.5 Å height by growing the interlayer at 1050 °C [Fig. 3(d)]. Additionally, by reducing the interlayer thickness as thin as 100 nm, re-enlargement of the hillock structures during the growth of the interlayer was minimized even though the interlayer was grown under relatively low T_g (1050 °C).

Zoom In Zoom Out Reset image size

As aforementioned, the density of the screw- and mixed-type FFA Sp-AlN dislocations was reduced when increasing the AlN film sputter-deposition temperature, which also reduced hillock density. Furthermore, the hillock structure size was reduced by controlling the MOVPE growth conditions. These results prove that we successfully improved the surface flatness of AlGaN on FFA Sp-AlN with a relatively small substrate off-cut angle of 0.2°. Lastly, the UV-C LED structures were fabricated on top of the smooth AlGaN film, and the optical properties of the LED were characterized. Regarding LED fabrication, FFA Sp-AlN was prepared with a 600 nm thick AlN film under the conditions of a sputtering temperature of 750 °C and thermal cycle annealing for 51 h at 1600 °C–1680 °C. The following layers were grown in the order presented: 600 nm thick AlN homoepitaxial layer, 100 nm thick compositionally graded UID-Al_1–0.77Ga_0–0.23N buffer layer, 1.2 μm thick Al_0.77Ga_0.23N:Si current-spreading layer with [Si] ∼ 1 × 10¹⁹ cm⁻³, three pairs of multiple quantum wells (MQWs) consisted of 2.0 nm thick Al_0.44Ga_0.56N well layers and 3.6 nm thick Al_0.69Ga_0.31N barrier layers grown at 1050 °C, 7 nm thick AlN electron-blocking layer, 15 nm thick compositionally graded Al_0.95–0.40Ga_0.05–0.60N:Mg p-type layer, and 150 nm thick GaN:Mg contact layer. The 100 nm thick top of the Al_0.77Ga_0.23N:Si current-spreading layer was grown at 1050 °C (low T_g interlayer) while the other underlying 1.1 μm thick part of the layer was grown at 1150 °C. Both of well and barrier layers of the MQWs were shallowly Si-doped ([Si] ∼ 3–5 × 10¹⁷ cm⁻³) to reduce the non-radiative recombination centers. ^28,29)

Figure 4 shows the crystalline quality of the UV-C LED epitaxial sample characterized via X-ray diffraction (XRD). The reciprocal space mapping (RSM) image obtained via (10$\bar{1}$5) diffraction [Fig. 4(a)] shows that the layers spanning the AlN template to the MQWs are coherently grown. The relaxation rate of Al_0.77Ga_0.23N:Si and MQWs against AlN are as low as 0.3% and 0.5%, respectively. Moreover, Fig. 4(b) shows the XRC profiles for the AlN template and Al_0.77Ga_0.23N:Si. The FWHM values for Al_0.77Ga_0.23N:Si were comparable with those for the AlN template for symmetric (0002) diffraction and asymmetric (20$\bar{2}$4) diffraction, indicating that the Al_0.77Ga_0.23N:Si retained the high crystallinity (low TDD) of the AlN template. The edge-type dislocation densities of the AlN template and Al_0.77Ga_0.23N:Si, estimated based on the XRC-FWHM values, were 1.1 × 10⁸ and 9.1 × 10⁷ cm⁻², respectively. Using the aforementioned procedures, the hillock density was determined to be 7.2 × 10³ cm⁻². These values are the state-of-the-art low TDDs among the Al(Ga)N grown on sapphire substrates.

Zoom In Zoom Out Reset image size

LED devices were fabricated using standard photolithography, inductively coupled plasma reactive ion etching, and electron-beam evaporation metallization. The p-type electrode was Ni/Au annealed at 600 °C for 5 min. The n-type electrode, GaN:Si, was selectively regrown on top of the exposed AlGaN:Si, atop which Ti/Al/Ni/Au was formed; this was followed by 500 °C annealing for 2 min. The sapphire substrate was thinned to 150 μm and polished, and then isolated into LED dies by applying laser scribing. The 400 × 300 μm² sized UV-C LED chip with a p-type electrode area of 9.8 × 10⁻⁵ cm² was flip-chip mounted with AuSn eutectic on a 6.8 × 6.8 mm² sized AlN-based container type ceramic package with Al reflector (t = 0.6 mm) on the top. No submount was used for the flip-chip bonding. Figure 5(a) shows the optical microscope image of the as-packaged UV-C LED bare die. To demonstrate the substantial potentiality of our devices, encapsulation with a UV-transparent silicone resin composed of poly(dimethylsiloxane) and poly(diethoxy-siloxane) ^30,31) was performed. The resin was potted on the die by hand to fabricate a hemisphere-like-shaped lens. Figures 5(b) and 5(c) show photographs of the LED packages without and with encapsulation. The light output power (LOP) was evaluated by mounting a Si photodiode (S1227-1010BQ, Hamamtsu photonics) directly on the LED package. The photocurrent was converted into LOP and EQE considering the spectra of the fabricated LED and the detector's wavelength-dependent sensitivity. In this work, angular-dependent distribution of the emitted light was not considered.

Zoom In Zoom Out Reset image size

Figure 5(d) shows the electroluminescence (EL) spectra of the fabricated UV-C LEDs. Single-peak emissions with a peak wavelength of 263 nm were obtained. Figure 5(e) shows the I–V–L characteristics and Fig. 5(f) shows the EQE results of the LEDs. The maximum EQE before the encapsulation was approximately 4.9% (at 2.0 mA), and the output power was 4.0 mW at 20 mA input. After the encapsulation with silicone resin, the maximum EQE was approximately 8.0% (at 2.9 mA), and the output power was 6.6 mW at a 20 mA input, indicating that the encapsulation increased the light extraction efficiency by approximately 1.6 times. The high EQE in the low-injection region verifies the outstanding IQE of the devices. Previously, we reported maximum IQE of 90% and 66% at room temperature (295 K) and 400 K, respectively from the 10 QWs on FFA Sp-AlN. ³²⁾ Actually, the detailed structure of the MQWs in this work is different from the previous one, we assume that the lower TDD and smoother surface morphology achieved in this work contributed to the higher IQE. Further, we speculate that the high EQE is owing to not only IQE but also to the improved current injection efficiency (CIE). The uniform thickness and AlN-molar fraction of the MQWs, AlN-EBL, and p-type AlGaN:Mg are attributable to the suppressed undesirable current crowding and overflow.

To the best of our knowledge, the developed UV-C LED has the highest EQE at a wavelength of approximately 265 nm, even considering the state-of-the-art devices fabricated on sapphire substrates and free-standing AlN substrates. ^{1–13,33,34)} This proves that FFA Sp-AlN is a promising substrate for UV-C LEDs and DUV optical applications.

In conclusion, a 263 nm wavelength UV-C LED was fabricated on an FFA Sp-AlN template. The conditions of FFA Sp-AlN fabrication and MOVPE AlGaN growth were optimized to reduce hillock density and size. The developed method improved the flatness of the AlGaN surface. Owing to the high crystallinity and superior surface flatness of AlN and AlGaN, a maximum EQE of approximately 4.9% was obtained for the UV-C LED bare die on the FFA Sp-AlN. As the substantial potentiality of the UV-C LED on the FFA Sp-AlN, a maximum EQE of approximately 8.0% was obtained after the encapsulation. Thus, this study demonstrated the potential of applying FFA Sp-AlN in optical DUV systems and paves the way for high-performing, cost-effective UV-C LED development.

Acknowledgments