LED Droop: Do Defects Play A Major Role?
Theorists are proposing that density activated defect recombination (DADR) can account for droop, the decline in a nitride LED’s external quantum efficiency at high drive currents.
The team from the University of Arizona, AZ, and the Phillips University of Marburg, Germany, reached this conclusion after modeling the recombination efficiency in the active region of light-emitting structures at various current densities. They found that the behavior of their model - which featured monolayer thickness variations, defects and compositional fluctuations - could fit efficiency vs current curves from real nitride devices emitting at 410 nm and 530 nm.
The origin of LED droop is highly controversial, and there have been claims led by engineers at Philips Lumileds and backed up by computational scientists at the University of California, Santa Barbara, that Auger recombination is to blame.
However, state of the art theoretical calculations by this US-German partnership suggest that Auger recombination rates in the nitrides are too small to explain the droop. And interband Auger recombination cannot account for droop either, according to this team, because it would require the energy separation between the higher and lower bulk bands to be very close to the fundamental band gap, which is generally not the case.
The members of the team at the Phillips University of Marburg are also considering whether phonon-assisted Auger recombination could account for droop.
“These calculations are numerically very intensive and the results depend far more critically on details of the used model and material parameters than for the usual direct Auger processes," explains Joerg Hader from the University of Arizona, who is the lead author on the paper.
He says that the initial results suggest that phonon-assisted Auger recombination rates are too low to account for LED droop.
Hader and his co-workers put forward DADR as a possible cause of LED droop after constructing a model where electrons and holes occupy states in local in-plane potential minima at low pump currents. These minima can be caused by fluctuations in either the composition of the well, or its width (see figure).
As the current is cranked up, carriers leave these local minima that are relatively free of defects. Some then reach non-radiative recombination centers, which prevent them from playing a role in light generation. This carrier capture explains the decline in light output at higher drive currents.
The team has shown that its DADR model, which assumes Auger recombination can be completely neglected, provides a good fit to experimental internal efficiency data for nitride active regions emitting at 410 nm and 530 nm.
To realize a good fit, the majority of carriers have to be located at energies less than 100 meV from the minimum. This implies fluctuations in the indium composition of 2 to 3 percent, which is entirely feasible.
One of the striking characteristics of LED droop is that it is more pronounced at longer wavelengths. Hader says that this trend is compatible with the researchers’ model.
“Longer wavelengths require deeper wells and therefore higher indium content in the wells," explains Hader. “This should lead to stronger disorder, more carrier localization and more defects – the basic ingredients for the DADR model."
One of Hader’s next goals is to finish a wide-bandgap version of the simulation software package, SimuLase, which is being developed by the company he spends some time working for, Nonlinear Control Strategies. “It will contain the DADR model as one of its tools."
He and his co-workers would also like to develop a better understanding of the general tendencies of both DADR and droop, such as their temperature dependence. To do this, the researchers need to develop a hopping-type model.
The researchers from the University of Arizona and the University of Marburg have modeled the radiative recombination in an InGaN active region featuring monolayer thickness variations and compositional fluctuations. As the current increases, carriers leave these potential minima and some end up at non-radiative recombination centers. That accounts for the decline in LED output at higher current densities.
Hader et al. Appl. Phys. Lett. 96 221106 (2010)