February 08, 2021 by Emmanuel Ikimi
LEDs may be a leading light source, but their traditional chip architectures are subject to a form of inefficiency known as ‘efficiency droop’. We look at how a new chip architecture developed by IBM, NIST (the National Institute of Standards), Rensselaer Polytechnic Institute, and the University of Maryland, has minimised such inefficiency.
The major limitation faced by traditional LED chip architectures is a phenomenon called ‘efficiency droop’. Efficiency droop is the corresponding drop in the internal quantum efficiency (IQE) of LEDs that comes with a rise in current density. Based on past research, some of the most frequently debated potential causes of efficiency droop is Auger recombination, junction heating effect, and density activated defect recombination. Even nanopillar LED design, which is reportedly one of the best in performance, is subject to such a weakness.
LEDs and their efficiency are an ongoing topic of interest to engineers and other researchers. Pictured: five LEDs are suspended on an electrical panel (namely a ‘breadboard') for inspection.
Image credit: Bigstock
Electronics Point has previously reported on a research team from Imperial College London who claimed to improve LED performance and lifetime. It is noteworthy that the research was unable to handle the challenge of efficiency droop. The team made good of their claim by placing a 2D nanoparticle layer between the transparent casing and the light-emitting semiconductor chip, thereby preventing damage to the chip and conserving the light emitted.
Unlike the research team, Babak Nikoobakht and his team used a different approach to achieve a comparatively better result by modifying LED chip architecture. It is a design that guarantees an up-to-1,000-times-order-of-magnitude rise in brightness from conventional submicron-sized LEDs.
The architecture includes GaN (gallium nitride), on which a lateral ZnO nanofin is epitaxially grown (as explained further below). The fact that ZnO, with its wurtzite crystal structure, has a similar bandgap to that of GaN: 3.36 electronvolts (eV) and 3.43 eV for ZnO and GaN respectively, but a comparatively greater exciton binding energy—namely 60 megaelectronvolts and 24 megaelectronvolts for ZnO and GaN respectively—makes it ideal for high-temperature lighting and optoelectronics
To elaborate on the epitaxial growth process, moreover, this involves about 5 microns-long ZnO fins, with spaces in between, that are grown on the p-type GaN material, leading to the formation of a comb-like structure. They produce light between violet and ultraviolet bands from facets exposed to air at an extraction efficiency of about 15%.
The research team took the following steps in fabricating the ZnO-GaN LED chip. Firstly, they applied the surface-directed Au (gold)-catalysed vapour-liquid-solid growth process in growing the fins by forming Au nanodroplets at the boundaries of the catalyst pattern.
ZnO fin grows laterally and uprightly on c-plane GaN, about 160 nanometres wide and five micrometres long. Subsequently, with the aid of photolithography and wet etch, the initial catalyst site is removed, which leads to the formation of isolated fins. These fins are often characterised by a couple of nonpolar side facets—one of which is coated by photolithography and angled-oxide deposition with a dielectric layer, leaving the other accessible. This accessible side facet receives an n-contact metal electrode.
A close-up of a multicolour LED strip
Image credit: Bigstock
To gain empirical evidence that this architecture works as expected, the wire-bonded fin LEDs were placed before a 44 mm diameter integrated sphere connected to a charge-coupled device array spectroradiometer, which measured each fin LED radiant fluxes at various drive currents. It was observed that at a total time of 325 seconds and as current increased from 1 to 10 milliamperes, two intense ultraviolet emission peaks emerge: one at 368.5 nanometres (3.369 eV) and the other at 378.5 nanometres (3.280 eV). It implies a continuous rise in intensity with a corresponding increase in current, which is the complete opposite of efficiency droop in a conventional LED chip architecture. It also signifies a suppression in GaN radiative recombination.
Additionally, with a total spectral radiant flux of a pixel that consists of five fins, the current was increased up to 100 milliamps, which is the maximum current used in the measurement setup explained above. For different current ranges and operation times, at a controlled 25 (±0.1)°C equilibrium temperature, a linear total radiant flux-current graph was generated, further reinforcing the observation that the fin LED chip increases its light intensity and output power at even higher current levels. A 100-fold increase in current resulted in a 108-fold increase in the optical power, implying an adequately controlled carrier loss. Although radiative recombination occurs at the ZnO fin, electron leakage, which causes the droop effect, is not observed in this fin LED architecture.
The team—comprising researchers from IBM, NIST (the National Institute of Standards), Rensselaer Polytechnic Institute, and the University of Maryland—further examined wall plug efficiency in LED and lasing modes to get a clearer view of recombination pathways in the fins. They used a voltage-current scan on the fins, from which they calculated the diver efficiency. This value was further useful to determine the internal quantum efficiency (IQE) of the fins. They observed that even at very high current densities of up to 1,000 kA/cm² (kiloamperes per square centimetres), the fin LED pixels did not show any sign of efficiency droop. Furthermore, they observed that when pixels attain a laser diode mode, the IQE rises sharply. This is noteworthy because conventional LEDs with increasing current density experience a sharp fall in their IQE.
A recent report observed that the conventional nanopillar LEDs reach 22-nanowatt power at a maximum external quantum efficiency (EQE) of 10-⁴ (ten to the power of negative four) and a current density of 100 kA/cm². At 40 times the current density, a competitive EQE of 6 by 10-5 is achievable as a fin LED architecture prevents EQE rollover, which results in pixels with over 1,000 times more output power than nanopillar LEDs.
The fin LED chip architecture holds a lot of potential in ensuring brighter LED lighting. Several methods were used by the researchers to show that the architecture had only a minimal efficiency droop. Their findings are a significant step that will encourage further research into efficient LED chip architectures that not only minimise the efficiency droop effect but also minimises ‘current density droop’ and ‘temperature droop’ effects in LEDs, therefore maximising the output power per pixel that the light-emitting diodes can achieve.
