Research review: Defects could account for LED droop's temperature dependency
A US-German collaboration has put forward more theoretical evidence to support its claim that density-activated defect recombination (DADR) could account for droop in GaN-based LEDs.
The team from Nonlinear Control Strategies, the University of Arizona and Philips University Marburg has included terms that can be responsible for droop – the decline in LED efficiency as the current is cranked up high – in calculations on device efficiency at temperatures from 100 K to 400 K. In general, in this regime the warmer the device gets, the weaker the droop causing mechanism becomes.
Reasonably good fits to experimental data are possible with the DADR model. With a model that accounts for droop with Auger recombination, however, it is only possible to replicate real data when the Auger recombination coefficient is about two orders of magnitude higher than the highest value produced by calculations. What’s more, the variations of the droop causing mechanism with temperature do not follow the generally assumed increase in Auger loss strength with temperature.
In other words, this recent theoretical work, which involves radiative loss calculations that include higher excitonic terms, indicates that Auger recombination is not the primary cause of LED droop. In the DADR model, at low current densities carriers are localized in regions with very few defects. But as the current increases, more and more carriers spill over into regions with strong defect recombination.
Energetic barriers separate these two regions, and carriers move between them via electron-electron scattering, which is more prevalent at high electron densities. Consequently, droop is stronger when the electron density is higher. “The strength of electron-electron scattering becomes weaker with increasing temperature,” says Joerg Hader from Nonlinear Control Strategies and the University of Arizona. This occurs because at higher temperatures there is a reduced probability of finding electrons with an energy close to the bandgap – more occupy higher energy states.
Hader and his co-workers have used their DADR model to fit graphs produced by other researchers, which detail LED efficiency at various drive currents and temperatures. Experimental results generated with a blue-emitting LED fabricated by Nichia and a green cousin made at Osram Opto Semiconductors were used in the theoretical study.
The DADR model contains two recombination processes involving defects: Schottky-Read-Hall (SRH) recombination, which is associated with crystal defects, such as threading dislocations; and DADR recombination.
“We don’t know what the defects are that are involved in DADR,” admits Hader. He says that they could be the same defects that are responsible for SRH recombination, but in this case they only contribute to droop at higher current densities, due to the energy barriers. Another possibility is that they are V-pits. “While we are not sure, they seem to fit the DADR picture.”
According to Hader, findings reported in a recent paper by Guan Sun and co-workers from Lehigh University (Appl. Phys. Lett. 99 081104 (2011)) support the DADR hypothesis. In that experimental work, the researchers uncover two decay times in InGaN/GaN quantum wells. One that is relatively slow and has a time scale typical for SRH recombination; and another that is faster, and has a time scale similar to that expected for DADR recombination.
“Moreover, through spectral analysis they conclude that the SRH-like decay originates from localised states and the fast decay happens at elevated densities when carriers start to occupy delocalised states,” says Hader.
Fred Schubert’s group at Rensselaer Polytechnic Institute, in collaboration with Samsung LED, published a paper in early 2011 detailing the temperature dependence of the strength of droop (D.S. Meyaard et al. Appl. Phys. Lett. 99 041112 (2011)).
According to Hader, that work focuses on LED behaviour above room temperature, which is a regime where carrier leakage is expected to become more significant as the device gets hotter. This expectation was found to be the case by this USKorean team.
“We don’t take that as a contradiction of our results,” says Hader, “since we look specifically at situations where leakage is not significant – low temperatures and lowto- medium currents.”
J. Hader et al. Appl. Phys Lett. 99 181127 (2011)
