UCSB simulations support Auger droop theory
Computational scientists at the University of California, Santa Barbara (UCSB), claim that Auger recombination contributes strongly to droop, the decline in the external quantum efficiency of GaN-based LEDs at higher drive currents (Appl. Phys. Lett 94 191109 (2009)).
The calculations by UCSB's Kris Delaney, Patrick Rinke and Chris Van de Walle show that the droop in visible LEDs is caused by interband Auger recombination, a form of non-radiative recombination that peaks at 2.5 eV (0.5 µm). This explains the origin of the "green gap", the decline in external quantum efficiency as the wavelength of GaInN LEDs is stretched from the blue to the green.
The UCSB team s result supports the controversial claims of LED manufacturer Philips Lumileds. Experimental studies by this Californian chip maker revealed Auger recombination, a non-radiative process involving the interaction between three carriers, including at least one electron and one hole, is to blame for the fall in GaN LED efficiency at higher drive currents.
But the conclusions of UCSB's simulations differ from those produced by theorists at the universities of Arizona and Marburg, who calculated that Auger recombination has a negligible impact on LED droop (see related story Simulations question Lumileds droop theory).
This discrepancy results from differences in the modeling of the nitride band structure - the UCSB researchers identified and included a second conduction band in their calculations.
"When we identified this second conduction band we were thrilled," says Rinke. "Then it was just a matter of working through the theory and putting the calculations of the Auger recombination rate on a rigorous footing."
Band structures for the nitrides were obtained by combining density-functional theory with many-body perturbation theory.
The researchers then calculated the Auger recombination rate. Explicit integration over a grid of points in momentum space was avoided, because this approach would be "prohibitively expensive". Instead, the researchers employed a Monte Carlo approach, which computed statistical averages over 40 million Monte Carlo steps.
"With the mechanism [of LED droop] now identified, future development can focus on removing or reducing losses due to Auger recombination," remarks Van de Walle.
Combatting droopThree approaches to diminishing these losses are discussed in the UCSB paper. Unfortunately, they all have their drawbacks.
One option is to produce nitrides LEDs with zinc-blende crystal structures, rather than wurtzite ones, because this pushes the second conduction band to far higher energies. But the growth of high-quality, phase-pure zinc blende nitrides is very challenging.
The alternative approaches employ strain engineering or tuning of InGaAlN composition to adjust the bandstructure. However, calculations show that any changes would be insufficient to provide a significant improvement in LED performance.
