Quantum tunneling boosts UV LED efficiency
Efficient, compact UV LEDs could serve many applications. These solid-state sources could be used for water purification, air disinfection and sensing, where they would replace bulky incumbent sources, such as those made from mercury.
Today, the UV LED is yet to fulfill this promise. Although commercial devices have been available for several years, there has been a low uptake of these sources. The primary barriers to widespread adoption are low efficiency and high cost "“ compared to the visible LED, efficiency is an order of magnitude lower, while cost can be more than a hundred times higher. Both inferiorities stem from fundamental differences between AlGaN, the ternary needed in UV LEDs, and InGaN, the alloy used for making blue and green emitters.
Our team at The Ohio State University is working to overcome these limitations by pioneering a novel device architecture. The emitters that we are developing are radically different from conventional UV LEDs, and employ a tunnel-junction to improve hole transport through the heterostructure. We believe that our device has the potential to slash the cost of the UV LED, while delivering a tremendous hike in its efficiency.
The lower efficiency of the UV LED, compared to its visible cousins, is highlighted in plots that graph external quantum efficiency and wall-plug efficiency as a function of wavelength (see Figure 1). For UV LEDs, the highest reported wall-plug efficiency value at 280 nm is only around 5 percent.
The root cause of this very modest efficiency can be uncovered by considering the factors that govern it: wall-plug efficiency is the product of internal quantum efficiency, electrical efficiency and light extraction efficiency.
Of these three, the internal quantum efficiency can now be a respectable 60 percent or more, thanks to efforts that have improved material quality. Carefully designed AlGaN-based buffer layers can control threading dislocation density and strain when UV LEDs are grown on non-native sapphire; while growth on native bulk AlN, grown by HVPE, offers a strong starting platform, thanks to a typical threading dislocation density of 103 cm-2.
Dragging wall-plug efficiency down to less than 5 percent is the electrical efficiency and light extraction efficiency. Both are related to poor hole injection within the device.
Figure 1. The highest reported external quantum efficiency (EQE) and wall-plug efficiency (WPE) values for UV LEDs.
The poor hole injection is caused by the low density of this carrier. It is a weakness that originates in the high activation energy of acceptors in high composition AlGaN "“ it is 0.62 eV in AlN, compared with just 0.14 eV in GaN. The upshot of the high activation energy is that it is almost impossible to make a direct p-contact to AlGaN in a conventional UV LED.
Popular approaches to overcoming this major limitation and improving hole injection are to add a p-GaN layer and to use a p-AlxGa1-xN/ AlyGa1-yN superlattice (both options are illustrated in Figure 2). These solutions have their downsides, however. Although using p-GaN to inject holes can cut resistance and increase electrical efficiency, the bandgap of GaN is less than that of the emitted light, so absorption losses rise and extraction efficiency falls. Turning to the p-AlGaN superlattice avoids impacting light extraction, due to the higher bandgap, but it leads to a high resistance that reduces electrical efficiency.
Figure 2. Switching from a conventional UV LED to a device that incorporates a tunnel junction leads to reduction in absorption and electrical losses.
The approach that we pursue, which is based on hole injection via a tunnel junction, has no major weaknesses because it does not lead to excessive absorption or electrical losses. The modification that we make to a conventional LED is to replace the hole injection layer with a UV transparent, conductive n-type AlGaN layer and a tunnel junction (see Figure 2). When designed correctly, the tunnel junction allows holes to "˜tunnel' into the p-AlGaN layer, an approach that enables low resistance and ultimately high electrical-injection efficiency. Thanks to the transparency of the n-AlGaN top contact, this architecture also cuts optical extraction losses.
When UV LEDs that feature tunnel junctions are forward biased, by applying a positive bias on the top contact, the top tunnel-junction layer is actually reverse biased. This results in interband tunneling and hole injection into p-AlGaN (see Figure 3 for an energy band diagram of the device). The hole current that is injected into the active region is identical to the tunneling current, which is controlled by the voltage drop across the reverse-biased tunnel junction. So as long as the tunnel junction is well designed, it will have a low tunneling resistance, and the voltage drop across the tunnel junction layer will be far lower than that across the UV LED active region.
Converting the promise of a UV tunnel-junction LED into a reality is by no means an easy feat, because it is very tricky to make tunnel junctions work in wide band gap AlGaN. Forming a tunnel diode typically involves degenerate doping of both the p+ and n+ regions of the junction to form a narrow depletion barrier. Current flow through the junction originates from the quantum-mechanical tunneling of carriers through the thin depletion barrier, and if the bandgap of this layer increases, the tunneling probability decreases exponentially. Consequently, tunneling resistance increases exponentially with increases in the barrier's band gap (see Figure 4).
Figure 3. The band diagram of a tunnel-junction UV LED under operation.
The major flaw with conventional AlGaN-based tunnel junctions that are suitable for UV LEDs is that they sport resistances in the 10-100 Ohm cm2 range. This would lead to voltage drops of 1000's of volts! Obviously, such high resistances are impractical. The good news, however, is that the III-nitride system provides a unique tool "“ polarization "“ to overcome these limits.
One of the key consequences of polarization is the high density of sheet charges at III-Nitride heterointerfaces. When a thin InGaN layer is inserted between p-type and n-type AlGaN layers, polarization charges at the AlGaN/ InGaN interface build up a high polarization field. This leads to dramatic band bending across the thin InGaN layer, so that the band edges of n-type and p-type AlGaN align over just a few nanometers (see Figure 3).
Inserting this thin layer of InGaN creates a triangular barrier that controls the interband tunneling. This is a bonus because this profile is more efficient for quantum mechanical tunneling. We combine p-Al0.3Ga0.7N with an In0.25Ga0.75N layer just 4 nm-thick and n-Al0.3Ga0.7N to create a tunnel junction along the metal-face orientation. This spawns net positive and negative polarization charges at the top and bottom AlGaN/InGaN hetero-interfaces, respectively (see Figure 5).
Figure 4. Reported tunnel-junction resistances as a function of band gap energy for different material systems.
Electrical measurements on our tunnel-junction UV LEDs reveal a sharp turn-on, with a voltage drop of 4.8 V at 20 A/cm2. Cranking up the current density to 2 kA/cm2 requires only 7.47 V (see Figure 6). We estimate that the resistance of the p-AlGaN/ InGaN/ n-AlGaN tunnel-junction layer is 5.6 à—10-4 Ω cm2, a figure that supports our view that polarization engineered tunnel junctions offer a low tunneling resistance for wide band gap materials.
Additional benefits that are demonstrated with these measurements are that the tunnel junction has a low resistance and introduces a low voltage drop, so it does not lead to a hike in LED power consumption. Another virtue of the tunnel-junction is that it cuts the overall device resistance, which trims Joule heating and increases device lifetime.
Figure 5 (a) High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image, (b) epitaxial stack, and (c) equilibrium energy band diagram of a tunnel-junction-based UV LED structure.
Figure 6: Current-voltage characteristics of the tunneling injected UV LED (50 μm à— 50 μm).
Figure 7 (a) Electroluminescence of a tunnel-junction UV LED structure with single peak emission at 327 nm. The inset shows an optical micrograph of a tunnel-junction UV LED device (50 μm x 50 μm) driven at 10 mA. There is partial top metal coverage, which appears as a dark region in the image. (b) External quantum efficiency and wall-plug efficiency of the device. Measurements are on-wafer, without using an integrating sphere.
Better contactsA further merit associated with incorporating a tunnel junction into the UV LED is that it allows a resistive, absorbing p-type metal contact to be replaced by one that is transparent, has a much lower contact resistance, and improves current spreading. Although the tunnel-junction introduces an absorption loss, this is estimated to be less than 4 percent, thanks to the InGaN layer being incredibly thin. In other words, photons can escape the top surface of our device with minimal absorption loss, enabling the demonstration of 327 nm devices with efficient, top-surface light emission (see Figure 7).
We have undertaken on-wafer power measurement on our devices. Results obtained without an integrating sphere reveal a maximum external quantum efficiency of 1.5 percent, and a wall-plug efficiency of 1.08 percent (see Figure 7). This indicates that the feeding efficiency "“ the ratio between the mean energy of the emitted photons and the voltage acquired by electron/ hole pairs "“ is 0.73. This high feeding efficiency is due to the voltage drop across the whole structure, which further confirms the benefit of using a tunneling contact over a conventional p-type contact. The low value for wall-plug efficiency is probably caused by the combination of an unoptimized active region design and a high dislocation density "“ it is in excess of 109 cm-2.
We believe it is possible to address both of these weaknesses and go on to produce UV LEDs that set a new benchmark for wall-plug efficiency and device operation. Once this is accomplished, it should open the door to the manufacture of highly efficient UV LEDs that serve a wide-range of commercial applications.
But that's not all "“ the tunnel-junction might be able to inject a new lease of life into other classes of emitter, via the creation of tunneling injected laser diodes, multi-color light sources, and LEDs with multiple active regions that deliver a hike in output power while cutting chip costs. So, in the years to come, don't be surprised if you hear more about the virtues of the tunnel-junction in optoelectronic devices.
UV LED efficiency: The benefit of the tunnel junction
Conventional UV LEDs are limited by: low electrical efficiency, due to a low thermal-activated hole density and poor p-type contacts; and poor light extraction efficiency resulting from high absorption loss and internal reflections. The wall-plug efficiency is limited to below 5.5 percent, and it can be even less than this at shorter wavelengths, due to more challenging hole injection problems.
In comparison, tunneling injected UV LEDs promise far higher wall-plug efficiencies "“ they could exceed 40 percent. Increases in efficiency over conventional UV LEDs result from a combination of high electrical efficiency that stems from non-equilibrium hole injection, and enhanced light extraction efficiency that is a result of the use of a transparent contact/spreading layer.
