UVB LEDs: Refining the design

LEDs emitting in the UVB are delivering an increase in efficiency, thanks to the introduction of grading to the hole-source layer and the electron blocking layer, and a thinning of the nickel film below the p-type contact.


BY M. AJMAL KHAN, NORITOSHI MAEDA, JOOSUN YUN, MASAFUMI JO AND HIDEKI HIRAYAMA FROM RIKEN AND YOICHI YAMADA FROM YAMAGUCHI UNIVERSITY

THERE IS MUCH MERIT in replacing a toxic, mercury-based ultraviolet light source with LEDs emitting in the same spectral range. Recently, LEDs emitting in the UVC, a domain below 280 nm, have attracted much attention, because they can kill Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), as well as disinfecting air, water, food, and surfaces. But there is also a great deal of opportunity for LEDs emitting in the neighbouring UVB band that is at slightly longer wavelengths. Narrow-band sources centred on 310 nm can be deployed for: cancer immunotherapy; the treatment of vulgaris, psoriasis, and atopic dermatitis; and plant growth with enriched phytochemicals (see Figure 1). In addition, emitters centred around the slightly shorter wavelength of 294 nm can be used to prevent plant diseases, attack the tomato mosaic virus, and produce vitamin D3 in the human body.

Figure 1. Some critical applications of UV light including UVB emission.

LEDs emitting in both the UVC and the UVB are based on AlGaN. This ternary nitride alloy has many promising features, providing reasonable optical and electrical properties, and the capability to form quantum wells with a high internal quantum efficiency. However, when the most common growth technology for making GaN-based LEDs is employed – that is low-pressure MOCVD –challenges are faced. When producing AlGaN-based epistructures on c-plane sapphire, it is far from easy to grow a highly conductive, silicon-doped n-AlGaN electron-source layer, and a highly transparent, magnesium-doped p-AlGaN hole-source layer. Difficulties arise from the large lattice-mismatch of 13.3 percent between c-plane sapphire and the AlN epilayer, as well as the possibility of kinetic separation when forming AlGaN layers with an aluminium composition of 40-45 percent.

At Riken, our team has addressed these issues by switching to a pulsed ammonia technique. This modification allows us to produce a high-crystalline-quality AlN template with a total dislocation density as low as 5 × 10⁸ cm⁻² on a c-sapphire substrate. We have been working with this technology for several years, using it to produce UVB LEDs. Unfortunately, even with these high-quality templates, devices have an external quantum efficiency of just 6.5 percent and a maximum light output power of 32 mW, according to room-temperature measurements on a bare wafer.

It is of no surprise that these devices are overshadowed by their blue cousins, but it is a concern that these results are inferior to those of UVC LEDs, which can realise external quantum efficiencies of up to 20 percent – although it’s worth noting that this superior result is supported by the introduction of micro-patterned sapphire, a flip-chip process, and expensive rhodium-like p-electrodes. A farer comparison is to use the result from a bare wafer, which provides an external quantum efficiency of around 8 percent at room-temperature. But even this suggests that the performance of UVB LEDs can be improved, a goal to which we aspire – and are making good progress.

Our recent breakthroughs have come through efforts on several fronts. Part of our success has stemmed from considering the relaxation condition in the n-AlGaN electron-source layer underneath the active region. The nature of this relaxation strongly influences piezoelectricity, extended defects, point defects, aluminium-alloy fluctuations and non-radiative recombination centres in the multi-quantum wells, which can ultimately degrade the internal quantum efficiency.

We have taken a step forward by realising a high-crystalline-quality in the AlGaN electron-source layer underneath the active region, where we reduced the density of total dislocations to 8 × 108 cm−2. Based on this figure, we can assume that the partially relaxed n-AlGaN electron-source layer has the potential to maintain this level of crystalline quality until the final p-AlGaN contact layer of the UVB LEDs. Confirming this assumption, we have found at Riken, together with Yamaguchi University, that when we insert a 3.4 µm-thick n-AlGaN blocking layer underneath an n-AlGaN electron-source layer with a relaxation ratio of about 49 percent, the density of total dislocations is as low as 7 × 10⁸ cm⁻² (see Figure 2 (a) to (e)). Thanks to this advance, the internal quantum efficiency in UVB active regions is as high as 80 percent, 64 percent and 54 percent, at temperatures of 100 K, 200 K and 300 K, respectively.

Figure 2. (a) Improvements to the performance of a 304 nm-band AlGaN-based UVB LED have come from grading the hole-source layer (HSL) and the electron-blocking layer (EBL). (b) Optical characterisation of the whole UVB LED (the p-AlGaN-based UVB LED crystal wafer is shown in the inset). (c) Atomic force microscopy image of the final p-AlGaN contact layer of a UVB LED. (d) High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of an LED sample. (e) Magnified HAADF-STEM image of an LED sample taken around the n-AlGaN electron-source layer (ESL) and the multiple-quantum well (MQW) region confirms a high-crystal-quality at the atomic level.

Another advance we have made is our development of highly transparent layers, having a relative transmittance of around 90 percent in the UVB. Such a high level of transparency has been realised in our aluminium-graded multi-quantum-barrier electron-blocking layer, our aluminium-graded hole-source layer, and our p-AlGaN contact-layer (see Figure 2(a) and (b)). When viewed with atomic resolution, we observe fine steps and terraces on the top p-AlGaN contact layer (see Figure 2 (c)).

In addition to the advances already discussed, we have made even more important breakthroughs by increasing the hole-injection efficiency and increasing the light extraction associated with the p-type electrode. Read on to discover the details of these valuable gains.

Ensuring higher hole injection
It is well-established that the performance of UVB LEDs do not benefit from a fixed composition in the multi-quantum-barrier electron-blocking layer and the AlGaN hole-source layer, as this does not ensure a conductive p-type terminal and high hole injection toward the active region. One weakness of this fixed-composition design is the substantial difference between electron and hole transport towards the active region, with highly energetic electrons from shallow donor levels easily overshooting the active region and ending up in the p-region of the device. Performance is also hampered by diffusion of unactivated magnesium atoms in the hole-source layer that head towards the active region through the undoped AlGaN final barrier. Due to these deficiencies, present in both UVB LEDs and laser diodes, efficiency droops under high current injection and poor hole injection toward the active region, resulting in relatively poor radiative recombination.

One way to improve the carrier transport in UV devices is to generate a three-dimensional hole level. This approach, pursued by a team from the University of Notre Dame, has led to UVC LEDs that feature a polarisation effect – this comes from adjusting the aluminium composition profile from the p-AlGaN hole-source layer to the opaque p-GaN contact layer. However, this approach cannot be directly applied to UVB LEDs, as it leads to an unsuitable, opaque p-GaN contact layer.

To ensure success, we have modified the design, shifting to a highly transparent pure aluminium-graded p-AlGaN hole-source layer that has an aluminium composition ranging from a high concentration to a lower concentration (see Figures 3 (a) and 3 (b)). More recently, we have improved on this, adopting a state-of-the-art strategy of using a moderately magnesium-doped, aluminium-graded p-type multi-quantum-barrier electron-blocking layer in our UVB LEDs. This design delivers three-dimensional hole generation and hole transportation via intra-band tunnelling from the aluminium-graded p-AlGaN hole-source layer toward the active region.

Figure 3. Secondary-ion mass spectrometry (SIMS) profiles of the aluminium and gallium composition in (a) an aluminium-graded p-Al0.58Ga0.42N p-AlGaN hole-source layer (HSL). The inset illustrates the polarisation-induced 3D hole generation in the aluminium-graded p-AlGaN HSL structure, and the bandgap diagram of the graded layer, where the negative polarisation charge field is created by an inclination of the aluminium-profile in the magnesium-doped p-AlGaN HSL. When deposition starts from gallium-face crystals and is graded from GaN to AlGaN, the polarisation bound charge is positive. This induces the generation of mobile three-dimensional electrons – it is a vice-versa situation for the generation of three-dimensional holes. The total polarisation, a combination of the spontaneous and the piezoelectric polarisation, may be pictured as charged dipoles in every unit cell of the crystal. Generating 3D holes in an aluminium-graded p-AlGaN HSL for UVB emitters is more challenging than it is in a hybrid structure of the aluminium-graded p-AlGaN HSL to p-GaN, due to the limited availability of the aluminium variation window to keep its transparency. When the composition of the layer is graded with an increasing aluminium mole fraction, the net unbalanced bound polarisation charge is negative. Alongside the [0001]-direction (gallium-polar), the aluminium composition in the p-AlxGa1-xN HSL gradually changes from a high aluminium content of 60 percent down to low aluminium content of 40 percent (see both (a) and (b)) and the polarisation interface charge is negative. Such polarisation-assisted interface negative charges induce a movable 3D hole gas, employed in the new LED design. (b) SIMS spectra of the aluminium and gallium composition in magnesium-doped p-multi-quantum-barrier (MQB) electron-blocking layer (EBL) including p-AlGaN HSL (new LED). (c) Estimated energy-band diagram of the newly designed p-AlGaN-based UVB LED (new LED). The aluminium composition of the p-AlxGa1-xN HSL of the UVB LED gradually decreases alongside the [0001]-direction (growth direction). This may enhance the 3D hole generation near the EBL-II of the p-MQB EBL.

The generation of our three-dimensional hole gas originates in the gallium-polar structure, which has a net negative charge at the interface between two layers of AlGaN with different compositions. Bound charges create a built-in electric field, leading to energy-band bending that would be greater than the bandgap of the semiconductor layer if left uncompensated. To neutralise these bound charges, negative polarisation charges and holes are field-ionised from neighbouring deep magnesium-acceptor atoms. This generates a high-density mobile 3D hole gas (see Figure 3 (a), which includes a more detailed explanation of this phenomena). By adopting this approach, we have increased hole concentration to around 3 × 10¹⁶ cm⁻³, reduced resistivity to typically 22 Ωcm, and realised a room-temperature hole mobility of 9.38 cm² V^-1 s^-1 in our moderately polarised hole-supplying layer, which has AlGaN graded from a content of 60 percent to 40 percent.

As well as improving the supply of holes, we have refined their transport to the active region by optimising the design of the electron-blocking layer. In previous work, we had introduced a pair of multi-quantum-barrier electron-blocking layers. They both had the same level of aluminium composition employed in our initial design. With our latest UVB LED we have switched to multi-quantum-barrier electron-blocking layers with a graded aluminium content (see Figure 2(a) and Figure 3(b)).

With this new design, the electron-blocking layer, which is nearer to the active region, has a slightly higher aluminium content than its sibling. This architecture ensures effective blocking of high-energy electrons while supporting hole injection through intra-band tunnelling after the generation of the two-dimensional gas at the interfaces between the electron-blocking layers and the ‘valley’ layer between them. By selecting a relatively small bandgap for the valley layer, holes are injected into this layer from the hole-source layer by thermionic emission and intra-band tunnelling, before being easily transported to the active region.

Figure 4. (a) The external-quantum efficiency of a 350 µm by 350 µm UVB LED emitting in the 304 nm-band improves with grading of the electron-blocking layer and the hole-source layer. (b) The light output rolls over in the reference design, but continues to climb in the superior variant.

To experimentally investigate the influence of grading the electron-blocking layers and hole-supplying layers, we have fabricated UVB LEDs, which feature a p-electrode of 1 nm nickel, followed by 200 nm of aluminium. We have compared this to our best reference LED, which includes electron-blocking layers and hole-supplying layers with a fixed aluminium content.

Our new device produces a single electroluminescence peak at 304 nm with a full-width at half-maximum of around 10 nm (see Figure 4 (a)). For our chip that’s roughly 350 µm by 350 µm in size, room-temperature bare-wafer measurements reveal a continuous-mode output power of 23 mW, up from 18 mW for the reference (see Figure 4 (b)). Additional encouraging results are a fall from 36 V to 23 V in the operating voltage for a 20 mA drive current, and an increase in maximum external quantum efficiency from 6.4 percent to 8.2 percent (see Figure 4 (a)).

Perfecting the p-electrode
Aluminium is known to be a good choice for the p-electrode, due to its high reflectivity. However, the underlying nickel film can compromise performance by absorbing UVB light and increasing the operating voltage. To address these issues, we have investigated devices with a thinner nickel layer, reducing this from 1-2 nm to just 0.4 nm, while employing an aluminium thickness of 200 nm. Devices have been fabricated with sizes of around 250 µm by 250 µm, 350 µm by 350 µm,
400 µm by 400 µm, and 450 µm by 450 µm (photos of the p-electrodes are shown in Figures 5 (a) and (b)).

Figure 5. Characterisation of UVB LEDs with a graded electron-blocking layer and the hole-source layer under continuous-wave (CW) and pulsed operation at room temperature (RT), using bare-wafer measurements. (a) Current versus output power (I–L), and current versus external quantum efficiency (I–EQE) (the inset shows an image of a real UVB LED with a chip size of 450 µm by 450 µm during operation (b). I–L and I–EQE characteristics, highlighting the benefit of a thinning of the nickel layer (an image of a real, small-sized UVB LED during operation is shown in the inset).

Room-temperature bare-wafer measurements on the smallest chip reveal that reducing the thickness of the nickel film leads to an increase in the external quantum efficiency from 7.8 percent to 9.6 percent. Driven at 18 mA, this chip produces a 5 mW output. For the biggest chips, the output is as high as 40 mW, under both continuous-wave and pulsed operation. We attribute these excellent results to a thinning of the nickel layer and the introduction of aluminium grading to the electron-blocking layer and the hole-source layer.

We have confirmed the benefits of thinning the nickel layer with optical modelling, which considers an aluminium thickness of 200 nm. Simulations show an increase in light extraction of 30 percent. At normal incidence, our new design with 0.4 nm-thick nickel has a reflectance of 85 percent, while that with 2 nm-thick nickel has a reflectance of just 76 percent. As well as the higher reflectance, a thinning of the nickel trims the contact resistance. According to our calculations, if our new design were accompanied by a standard UVB LED package, a nano-patterned sapphire substrate, lenses, a photonic crystal structure, a rhodium-based p-electrode, a heat sink, and a flip-chip architecture – all have been adopted for UVC LEDs – then an external quantum efficiency of 21 percent should follow.

Additional opportunities
The external-quantum efficiency is the product of three factors: the internal quantum efficiency, the carrier-injection efficiency, and the light-extraction efficiency. We have plans for improving all three.

To increase the carrier-injection efficiency, we know that we need a hole concentration of approximately 3 × 10¹⁸ cm^-3 and shallow acceptor energy levels in the p-AlGaN hole-source layer (see Figure 6 (a)). We have tried to tackle this challenge by applying an excimer laser anneal to a lightly polarized p-AlGaN hole-source layer. This has resulted in a hole concentration of approximately 2.6 ×10¹⁶ cm^-3, a hole mobility of around 9.6 cm² V^-1 s^-1 and a resistivity of 24.39 Ω-cm. While these figures are encouraging, such a p-AlGaN hole-source layer is incapable of matching the number of holes to the number of electrons that are delivered to the active region.

What we need to do to bolster hole injection is to switch to a new heterostructure with a new wide-bandgap material that has a highly activated p-dopant. Our replacement must satisfy three criteria: it must realise epitaxial growth on AlGaN and AlN; it must have a bandgap profile that aligns with AlGaN and AlN, to assist the blocking of high-energy electrons and promote hole-injection toward the active region; and it must have a lower acceptor energy level and a higher hole density than the aluminium-rich AlGaN hole-source layer (see Figure 6 (b)).

Figure 6. (a) Hole concentration versus aluminium composition in the p-AlGaN hole-source layer (HSL). (b) An experimental roadmap to grow p-type hexagonal (h) BN on a p-AlGaN HSL (Step-I) on AlGaN multiple quantum wells (MQWs) of UVB LEDs, and assisted by excimer laser annealing (ELA) treatment. Step-II is to grow highly efficient AlGaN superlattice-based UVB LEDs on epitaxial layer overgrowth (ELO) AlN/striped c-sapphire substrates.

Satisfying all these conditions is hexagonal p-type BN. According to theoretical considerations, if we can place impurity dopants at the interstitial sites in hexagonal BN, this will create a shallow acceptor energy level of 30 meV with high hole generation in the valence band. Due to this promise, we are planning to fabricate a new form of UVB LED that features p-type hexagonal BN and a polarized p-AlGaN hole-source layer, formed by treating with an excimer laser to activate holes. We are aware that it will be challenging to produce this structure by low-pressure MOCVD at a high growth temperature.

For the internal quantum efficiency and the light extraction efficiency, we are targeting values in excess of 70 percent and 17 percent, respectively. One optimistic idea of ours for reaching these milestones is to grow a highly transparent p-AlGaN hole-source layer and an epitaxial-layer-overgrowth AlN template on a striped c-plane sapphire substrate.

Such a device could be produced by growing an AlGaN-based UVB LED on (0001)-oriented epitaxially laterally overgrown AlN on a striped c-plane sapphire substrate. We have already shown that AlN formed by this approach has a threading dislocation density of just 1 × 10⁶ cm⁻², suggesting that the active region could have an internal quantum efficiency as high as 80 percent.

Another idea of ours is to realise an even lower threading dislocation density in the active region by turning to an AlGaN superlattice buffer layer on (0001)-oriented epitaxial-overgrown AlN on striped c-plane sapphire. This approach should prevent threading dislocations from penetrating through the AlN buffer (see Figure 6 (b)).

Improvements can also come from introducing the standard flip-chip package, featuring a heat management module, hermetic sealing and a specially designed lens for light concentration. We have made much progress with our compositional grading of hole-source layers and electron-blocking layers, and the thinning of the nickel film – but there is still much more to explore, as we continue to refine the design of the UVB LED.


Further reading
† M.A. Khan et al. Achieving 9.6% efficiency in 304 nm p-AlGaN UVB LED via increasing the holes injection and light reflectance. Sci Rep 12 2591 (2022)
† M.A. Khan et al. External Quantum Efficiency of 6.5% at 300 nm Emission and 4.7% at 310 nm Emission on Bare Wafer of AlGaN-Based UVB LEDs. ACS Appl. Electron. Mater. 2 (2020)
† M.A. Khan et al. Impact of Mg level on lattice relaxation in a p-AlGaN hole-source layer and attempting excimer laser annealing on p-AlGaN HSL of UVB emitters. Nanotechnology 32 055702 (2021)