Simulations question Lumileds' droop theory
Theorists from the universities of Arizona and Marburg, Germany, are challenging Philips Lumileds' claim that Auger recombination is the major cause for LED "droop", the decline in emission efficacy at higher drive currents.
Simulations by the US-German partnership show that Auger recombination - a non-radiative process that involves the transfer of energy from electron-hole recombination to a third carrier "“ has negligible impact on LED droop (Appl. Phys. Lett. 92 261103 (2008)).
This latest paper, which is co-authored by experimentalists at Osram Opto Semiconductors, also says that resonant optical excitation experiments show that droop cannot be caused by carrier leakage exclusively. This is a view that Lumileds shares, but is not held by all researchers in this field.
Mike Krames, who is the leader of the team that proposed the Auger theory (see related story An answer to LED droop?), told Compound Semiconductor that he actually welcomed the publication of these simulations. He says that this shows that the scientific community is shifting towards an explanation of droop based on fundamental phenomena and away from theories involving dislocations, InN fluctuations, and current-transport effects.
Krames also believes that these simulations do not disprove Lumileds' explanation for LED droop. According to him, the models that the theorists used do not account for higher energy bands in the material, and they only deal with one form of Auger recombination, "direct" band-to-band processes.
Compound Semiconductor put these points to Jorg Hader, one of the theorists at the University of Arizona.
Hader agrees that the calculations fail to account for higher energy bands, but he believes this would only alter the overall contribution by Auger recombination by tens of percent.
"Even if our calculations were off by much more than what I can imagine "“ say a factor of ten "“ the resulting Auger rates would still be a couple of orders of magnitude too small to explain the observed droop."
Complex calculationsHader explains that the importance of the Auger process in LEDs is not obvious, and it can't be measured experimentally. "It's very complex to treat the problem thoroughly, so people are speculating all the time."
Some researchers believe that Auger recombination is relatively small in the nitrides, due to their wide bandgap. But others argue that the Coloumb interaction between carriers in this material is very strong, leading to a high probability for Auger recombination.
To get to the bottom of all this, Hader and his coworkers employed sophisticated models that calculate "direct" band-to-band processes.
These simulations do not use uncontrolled approximations, such as averaging over spin or momentum indices or simplifications to the Coulomb matrix elements. This is claimed to reduce the uncertainty of the results to less than about 20 percent.
A thorough approachHader says that the formula used to solve Auger recombination is not new "“ in fact, it's been around for 40 years, but is so complex that accurate simulations are very difficult to run.
He says that many researchers have tried to take short cuts, but this has led to uncontrollable errors.
To avoid repeating these errors, Hader and his co-workers employed precise simulations, even though they have to pay the penalty of greater number-crunching. "Nowadays we've got it down to 10 minutes, but when I started and had the first version of the code ready it would have taken two to three years to run."
One of the difficulties with this calculation is associated with the uncertainties in the internal electric field strength, and the indium profile across the well. "We covered the whole space to make sure that the uncertainty in those parameters didn't affect Auger loss," explains Hader.
Ruling out Auger loss, carrier leakage and non-capture of carriers as reasons behind droop has led Hader and co-workers to speculate more exotic intrinsic loss mechanisms. According to them, two possible candidates are defect-assisted Auger recombination and phonon-assisted Auger recombination.
"I personally do not think that phonon or defect-assisted Auger processes are the reason for the droop, but I have no real means of discarding them," reveals Hader.
Collaborators at Marburg are writing the code to investigate the role of phonon-assisted processes in GaN LEDs.
Meanwhile, Hader is devoting some of his time to a spin-off company, Nonlinear Control Strategies Inc., that is commercializing the simulation code into software that will allow other researchers to perform similar calculations.
