Technical Insight
LED droop: Direct Auger gets the blame
Direct Auger recombination causes droop, but its impact can be diminished inserting graded layers into quantum wells to smooth the confining potentials.
Calculations by a partnership between at Technion-Israel Institute of Technology and the US Naval Research Laboratory show that a 'softer' potential reduces Auger recombination, leading to an increase in LED efficiency.
The debate over the origin of LED droop has taken yet another twist, with a partnership between researchers at Technion-Israel Institute of Technology and the US Naval Research Laboratory claiming that direct Auger recombination is to blame.
This view is at odds with that of many other theorists, who account for droop - the reduction in light efficiency as current is cranked up - with either more complex forms of Auger recombination or models involving defects.
For example, Chris Van de Walle's computational science group at the University of California, Santa Barbara, claims that the primary causes of droop are Auger-related processes involving phonons and alloy disorder. And a partnership between researchers at Boston University and Politecnico di Torino, Italy, argue that although Auger-related processes contribute to droop, carrier leakage, compositional fluctuations and threading dislocations may also play significant roles.
These differences over the cause of droop stem from differences in the structures under study, according to Roman Vaxenburg from Technion-Israel Institute of Technology. He has his co-workers have looked at quantum well structures, while others have focused on calculations for bulk material.
"In general, in bulk material, Auger recombination is not efficient due to strict energy and momentum conservation requirements," explains Vaxenburg. "On the contrary, in quantum-confined systems, such as quantum wells, the momentum conservation requirement if lifted and Auger processes are enhanced." Values calculated by the US-Israeli team for the Auger coefficient in quantum wells are in the range 10-31- 10-30cm6s-1.
The structures studied by this team feature symmetric quantum wells, which are found in non-polar LEDs. That does not mean, however, that their findings offer no insights into the vast majority of devices made today, which are polar LEDs grown on c-plane sapphire. "We expect that the Auger rate will be further accelerated in quantum wells grown in the polar direction," says Vaxenburg.
For their calculations, he and his co-workers use the well-established Pidgeon and Brown model which was proposed in the 1960s. This is often used for calculating the characteristics of direct bandgap semiconductors with a band-edge at the Gamma point of the Brillouin zone.
"It takes into account the eight band-edge sub-bands," explains Vaxenburg. "Adding more bands will only accelerate the rate of Auger recombination, because it will increase the density of final states."
The code employed for these calculations has been written from scratch. It is a few tens of thousands of lines long, and when it is run on a powerful personal computer, it takes 8-10 hours to calculate the total Auger recombination rate for a given quantum well.
Auger rates have been calculated in a wide variety of InGaN/GaN quantum well structures with different confining potentials. One class of structures features a conventional active region, with In0.25Ga0.75N barriers surrounding a GaN quantum well with a thickness of either 1.5 nm, 2.0 nm, 2.2 nm or 2.5 nm. Other types of structure under study feature 'softer' potentials, resulting from the insertion of layers with intermediate compositions.
The conclusions of the team should not be questioned due to their assumption of zero temperature. "At temperatures higher than zero, the Auger rate is expected to accelerate even further," says Vaxenburg.
Calculations reveal that using 11 layers to form a quantum well, rather than the conventional number, 3, can lead to a three-fold reduction in the Auger recombination rate. This is claimed to spur a 20 percent increase in LED efficiency, shift droop onset to a higher current density, and reduce the droop development rate at higher current densities.
The next goal for the team is to extend their model to include piezolelectric polarization.
R. Vaxenburg et al. App. Phys. Lett. 102 031120 (2013)
