Tunnel-injected sub 290 nm ultra-violet light emitting diodes with 2.8% external quantum efficiency

III-Nitride deep ultra-violet light emitting diodes (UV LEDs) have attracted great research interest because of a large variety of applications, including water purification, sterilization, and medical sensing.¹ Significant improvements in AlGaN material quality have been achieved based on the optimization of growth techniques, and this has led to the demonstration of active regions with high radiative efficiency.^2–8 However, in order to enable p-type ohmic contact and electrical injection, a heavily doped p-GaN layer is typically adopted in the conventional UV LED structure. This results in severe internal light absorption and poor light extraction efficiency. Moreover, the extremely low thermally activated hole concentration in ultra-wide bandgap p-AlGaN layers leads to poor electrical injection efficiency. Therefore, AlGaN-based deep UV LED efficiencies remain significantly lower than that of their InGaN-based blue LED counterparts.^1,4

Recently, tunnel-injected UV LED structures were proposed to solve both the absorption and electrical loss issues faced by the conventional UV LED structures.⁹ In this structure, a transparent n-AlGaN top contact layer is connected to the p-AlGaN layer through an abrupt tunnel junction layer. This avoids the absorbing p-type contact layers and enables efficient non-equilibrium hole injection.^10–12 Since the structure is terminated with an n-AlGaN layer, which has low spreading resistance, the top metal electrode coverage can be minimized to allow light extraction from the top surface.¹³ Besides, aluminum, as the only UV reflective metal with a high reflectivity above 90% in the deep UV range, can be used to form n-type reflective top contact.^14–16 This provides flexibilities in the development of optimized light extraction schemes based on the tunnel-injected UV LED structure. Another benefit of the tunnel-injected UV LED structure is that holes are injected through interband tunneling, which resists the influence of the extremely low thermal activation rate of Mg acceptors in AlGaN.^17,18 This is critical for deep UV LEDs, as the thermally activated hole concentration decreases significantly with the increasing Al content in the p-AlGaN layer. While conventional UV LEDs have encountered a substantial efficiency reduction when lowering the emission wavelength in the deep UV wavelength range, the tunnel-injected UV LED structure could potently solve the problem and lead to efficient deep UV emitters.

Low resistance (<1 mΩ cm²) GaN tunnel junctions have been achieved by polarization engineering^19–21 or degenerate doping.^22–25 The incorporation of GaN tunnel junctions has led to successful demonstrations of blue LEDs^10,24–26 with a wall-plug efficiency higher than 70%,¹⁴ cascaded LEDs,^11,27,28 edge emitting laser diodes,^29,30 and vertical-cavity surface-emitting laser diodes.²³ However, the fabrication of low resistance and low voltage consumption tunnel junctions for ultra-wide bandgap AlGaN is challenging because of the wide depletion barrier and doping limitations. Recently, polarization engineering was used to significantly reduce the interband tunnel barrier.⁹ Tunneling hole injection into p-AlGaN was demonstrated with appreciable light emission in a wide UV wavelength range of 325–257 nm from the tunnel-injected UV LED structure.^{9,13,16–18,31} Incorporation of tunnel junctions into nanowire LEDs has also been reported with excellent device performances.^32,33 In this work, we demonstrate efficient deep UV LEDs with an on-wafer peak external quantum efficiency of 2.8% achieved through non-equilibrium tunneling hole injection into p-Al_0.65Ga_0.35N.

The tunnel-injected UV LED structure was grown by N₂ plasma assisted molecular beam epitaxy (MBE) on a metal-polar Al_0.7Ga_0.3N template. The threading dislocation density in the substrate was estimated as 3 × 10⁹ cm⁻². As shown in Fig. 1(a), the MBE growth was initiated with 600 nm n⁺-Al_0.65Ga_0.35N as a bottom contact layer, this was followed by a 50 nm n-Al_0.65Ga_0.35N cladding layer, a quantum well (QW) active region, an 18 nm p-AlN electron blocking layer (EBL), 35 nm p-Al_0.65Ga_0.35N, a tunnel junction layer, 200 nm n⁺⁺-Al_0.65Ga_0.35N, and a 40 nm reverse graded n⁺⁺-AlGaN top contact layer with an Al mole fraction grading from 65% to 15%. The QW active region has an asymmetric quantum barrier design to account for the drift mobility difference between electrons and holes, with three periods of 2 nm Al_0.5Ga_0.5N QWs separated by 6 nm Al_0.65Ga_0.35N barriers, while a 3 nm Al_0.65Ga_0.35N barrier was adopted right below the EBL. The tunnel junction layer consists of p⁺-AlGaN with [Mg] = 5 × 10¹⁹ cm⁻³, 4 nm In_0.2Ga_0.8N, and 5 nm n⁺-AlGaN ([Si] = 1 × 10²⁰ cm⁻³) with the Al composition grading from 58% to 65%. The simulated energy band diagram confirms a sharp band alignment obtained through such a tunnel junction design. Meanwhile, the reverse-graded top contact layer provides a flat conduction band for electron transport [Fig. 1(c)] from the tunnel junction and enables the formation of a low resistance Ohmic contact without the necessity of metal annealing.³⁴ Figure 1(b) shows the high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) image of the tunnel-injected UV LED structure, where abrupt heterointerfaces are observed. No sign of lattice relaxation was observed for the AlGaN/InGaN tunnel junction layer from the STEM images.

Device mesa was defined by etching to the bottom contact layer using inductively coupled plasma reactive ion etching (ICP-RIE). The V(20 nm)/Al(80 nm)/Ti(40 nm)/Au(100 nm) metal stack was evaporated and annealed at 860 °C for 4 min to form the bottom contact.^18,35 The top contact was formed using an Al(30 nm)/Ni(30 nm)/Au(200 nm)/Ni(20 nm) non-alloyed metal stack. The top contact metal covered 37% of the 30 × 30 μm² device mesa area. To minimize internal light absorption, a low power ICP-RIE etch was then used to remove the reverse graded top contact layer in the area without the top metal contact as shown in the inset of Fig. 1(c). Then, electrical and optical characterizations were carried out for the on-wafer devices.

The bottom contacts and top contacts were characterized using circular transfer length measurement (CTLM) and TLM measurements, respectively. By optimizing the annealing conditions, we obtained near-Ohmic contacts on the n⁺-Al_0.65Ga_0.35N bottom contact layer as shown in Fig. 2(a). The resistances are extracted from the linear portion of the CTLM curves and are plotted with gap spacing after correction. The extracted bottom contact specific resistance is 1.2 × 10⁻⁵Ω cm². In comparison, the Al-based non-alloyed top contact provides a lower contact resistance of 2.2 × 10⁻⁶Ω cm² to the n⁺⁺-Al_0.65Ga_0.35N top contact layer. This is expected to originate from the lowering of the Schottky barrier formed at the metal-AlGaN interface and the formation of a flat conduction band profile in the heavily n-type doped reverse-graded top contact layer.¹⁸ The sheet resistances are extracted to be 411 and 216 Ω/◻ for the bottom and top Al_0.65Ga_0.35N contact layers, respectively. This ensures sufficient current spreading in the contact layers. The adoption of the graded contact layer provides a non-destructive method to enable low resistance contacts to the n-AlGaN top contact layer, as opposed to the conventional method utilizing alloyed contacts annealed at high temperatures, which could lead to severe shunt leakage due to metal spike into the active region.

The low contact resistances enable the analysis of the tunnel-junction resistances. The current-voltage (IV) characteristics of the 30 × 30 μm² tunnel-injected UV LED device are shown in Fig. 3. When negative voltage is applied on the top contact, the active region is reverse biased, and low leakage current is measured. By applying positive bias on the top contact, holes can be injected through the reverse biased tunnel junction layer into the forward biased active region. For ideal tunnel junctions, efficient interband tunneling should happen at low reverse bias. However, the tunnel-injected UV LED device showed a high voltage drop of 10.5 V at 20 A/cm². This is 5.9 V higher than the ideal turn-on voltage determined from the quantum well bandgap (E_g = 4.6 V). The excess voltage drop could have contributions from both the tunnel junction layer and the electron blocking layer. Previous demonstrations have shown that background compensation in the p-AlGaN layer could lead to an extended depletion barrier and therefore contribute to the increased voltage drop across the tunnel junction layer for efficient tunneling.³¹ 

The device reached a high current level of 1 kA/cm² at 15.6 V with a low differential resistance of 1.9× 10⁻³Ω cm². This can be estimated as the tunnel junction resistance by neglecting the contributions from the spreading resistances, contact resistances, and the p-n junction series resistances. Figure 3(b) summaries III-Nitride tunnel junction resistances (extracted at 1 kA/cm²) and excess voltage drop (extracted at 20 A/cm²) obtained from tunnel-injected III-Nitride LEDs.^9,17,18,31 Low tunnel junction resistances below 2 × 10⁻³Ω cm² have been achieved for ultra-wide bandgap AlGaN with the Al composition up to 75%.¹⁸ This makes it promising in the applications of tunnel-injected UV LEDs and laser diodes. However, a substantial increase in the excess voltage drop is observed as the AlGaN bandgap increases in the tunnel junction structure. Even though polarization engineering has been utilized to shrink the tunnel barrier through the insertion of an ultra-thin InGaN layer between p⁺- and n⁺-AlGaN layers, the large conduction band and valence band offsets at the hetero-interfaces lead to extended depletion barriers as shown in the band diagram in Fig. 1(c). As the Al content increases in the AlGaN layers, higher depletion barriers and reduced tunneling probability are resulted. Therefore, further optimizations in the tunnel junction layer are necessary to reduce the excess voltage drop.

Figure 4(a) shows the electroluminescence (EL) emission spectra measured from the top surface of the tunnel-injected UV LED device (30 × 30 μm²). Efficient light emission at 287 nm with a full width at half maximum of ∼12 nm was obtained. The inset shows the logarithmic plot of an EL spectrum. No parasitic emission was observed except for a weak emission peak at 510 nm, which was also detected in the tunnel-injected UV-A LEDs emitting at 325 nm.¹³ This could come from defect emission or light emission from the thin InGaN layer due to photoexcitation. The microscope image shows uniform light emission from the whole mesa region even though only 37% of the mesa region was covered by the top metal contact. While the reverse graded AlGaN contact layer was etched in most of the mesa area and did not contribute to light absorption, inevitable absorption persisted below the top metal contacts. Further optimizations in reducing the metal contact area while maintaining sufficient current spreading is necessary to maximize the light extraction efficiency.

The output power was measured on-wafer under continuous wave (CW) operation. The measured power values are normalized to the device mesa area and are plotted in Fig. 4(b). A high power density of 54.4 W/cm² was obtained at 1 kA/cm². The peak external quantum efficiency of 2.8% and a wall-plug efficiency of 1.1% were measured at 30 A/cm². Even though the measured power and efficiency values are underestimated by the on-wafer measurement, the efficiency values are comparable to state-of-the-art device results obtained for conventional UV LEDs.^36,37 Better device performance is expected by introducing surface patterning and device packaging for improved light extraction efficiency. Nonetheless, this preliminary demonstration confirms that efficient hole injection into ultra-wide bandgap p-AlGaN can be achieved through interband tunneling. However, the wall-plug efficiency is still limited by the high turn-on voltage. Optimization of the tunnel junction layer to reduce the extra voltage drop is critical in achieving high wall-plug efficiency.

In summary, we have demonstrated a high efficiency tunnel-injected deep UV LED employing a polarization engineered p⁺-AlGaN/InGaN/n⁺-AlGaN tunnel junction structure. By introducing a reverse compositionally graded n⁺⁺-AlGaN top contact layer, a low resistance contact (ρ_c = 2.2 × 10⁻⁶Ω cm²) was formed to ultra-wide bandgap n⁺-Al_0.65Ga_0.35N without the necessity of metal annealing. From the IV characteristics of the device, a low tunnel junction resistance of 1.9 × 10⁻³Ω cm² was extracted at 1 kA/cm². However, the device showed a high forward voltage drop of 10.5 V at 20 A/cm², indicating a large depletion barrier for tunneling. Non-equilibrium hole injection was confirmed by efficient light emission at 287 nm. The measured on-wafer peak external quantum efficiency and wall-plug efficiency are 2.8% and 1.1%, respectively. At 1 kA/cm², a high power density of 54.4 W/cm² was measured under CW operation. These results demonstrate the feasibility of achieving efficient hole injection into ultra-wide bandgap AlGaN and point to a pathway towards high efficiency compact ultraviolet emitters.

We acknowledge funding from the National Science Foundation (Nos. ECCS-1408416 and PFI AIR-TT 1640700) and the OSU TCO Accelerator Award. J.J., G.C., and J.H. acknowledge support from the Air Force Office of Scientific Research (AFOSR) under Contract No. FA9550-17-1-0227 and the Office of Naval Research (ONR Grant No. N00014-15-1-2363, Program Manager: Dr. Paul Maki). Sandia National Laboratories is a multimission laboratory managed and operated by the National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy's National Nuclear Security Administration under contract No. DE-NA-0003525.