A mix of modelling and experiment creates a strong case for carrier asymmetry as a primary cause of LED droop

 



CARRIER asymmetry has been added to the growing list of conjectures to account for droop, the decline in the efficiency of GaN-based LEDs at high current densities. This potential cause of droop is being put forward by a team from Rensselaer Polytechnic Institute, New York, and Samsung LED, Korea. Their claim for asymmetry as the cause of droop followed modelling device behaviour in p-n junctions and measurements of LED output at various carrier densities and temperatures.

The researchers also concluded that LED droop starts to kick in when conductivity in the p-side due to electron injection equals the conductivity due to the holes present in that layer. This condition, which the team refers to as high-level injection, creates an electric field on the p-side that increases leakage and droop.

Two factors are responsible for the severe carrier asymmetry in GaN LEDs: A very high ionization energy of the common hole dopant, magnesium, which leads to a low hole concentration; and a far lower mobility for holes than electrons.

The team studied the effects of this asymmetry in a very simple structure, a GaN p-n homojunction, using the commercial software APSYS.

They found that high-level injection produces: A voltage drop that is not confined to the depletion region but extends to a quasi-neutral region; and an LED series resistance that leads to deviations from the expected exponential current-voltage behaviour.

What’s more, the team uncovered theoretical evidence for the dependence of the recombination location on carrier symmetry. When a junction had equal doping on both sides, the recombination location is shifted just 1.6 nm to the p-side; but this rises to 18.3 nm when n-doping is far higher than p-doping.

To study droop in real devices, the team measured the LED’s external quantum efficiency. Unpackaged 1 mm by 1 mm thin-film chips that were grown by MOCVD – which featured a typical AlGaN electron blocking layer and had an exposed nitrogen-face that was roughened to increase light extraction – were driven at currents from 0.01 mA to over 1000 mA at temperatures ranging from 80 K to 450 K. Devices were driven in a manner designed to minimize self-heating. At 80 K, LED droop kicks in at a very low drive current and is more severe than it is at higher temperatures. That’s because the asymmetry in carrier concentration is relatively high, due to the incredibly low free hole concentration at this low temperature. “In addition to this, Shockley-Read-Hall recombination is minimized at 80 K, so the peak efficiency at very low currents is very high, making droop effects more easily recognized,” says Fred Schubert from Rensselaer Polytechnic Institute.

Measurements of the LED’s currentvoltage characteristics reveal a deviation from exponential behaviour. This occurs at the onset of high-level injection, which takes place when droop starts to kick-in. The team argues that this high-level injection produces a build-up of electric field in the p-type region, leading to enhanced electron leakage and a shift in the recombination region into the p-side. Schubert has previously argued that droop is caused by polarization fields, which lead to electron leakage out of the active region. “The common link between these two claims – asymmetry and polarization fields – is leakage of electrons over the EBL or, equivalently, lack of hole injection into the active region,” says Schubert. According to him, electron leakage and a lack of hole injection may be viewed as two sides of the same coin, because every hole that is not injected will lead to the leaking out of an electron.

“We believe that polarization fields play a significant role in c-plane GaN LEDs, including their role in letting electrons leak out of the active region,” says Schubert. “However, our and others’ efforts in reducing the polarization fields – by polarization matching and non-polar growth – have not reduced the droop to zero.” For this reason, Schubert and his co-workers have concluded that there must be another cause for LED droop: Carrier asymmetry.

Another popular conjecture for LED droop is Auger recombination, a non-radiative process involving three carriers. Researchers in this camp are trying to pin down the exact form of Auger recombination, with mechanisms involving phonons and alloy-disorder in the quantum wells receiving recent support. “Due to the very low carrier concentrations at which droop occurs, as well as the temperature dependence of the droop behaviour, we believe that even these newly-proposed forms of Auger recombination are very unlikely to cause significant droop in GaN/GaInN LEDs,” argues Schubert.

Another conjecture for droop, which has been proposed by Joerg Hader from the University of Arizona, is density-activated defect recombination (see page 53).

Schubert can’t rule this out, but he points out that in a recent paper by Hader and his co-workers, those researchers revealed that density-activated defect recombination overestimates the internal quantum efficiency at very high currents, which is probably due to carrier leakage and noncapture. Schubert’s team is now working on complementing its model with analytical expressions to predict the conditions for when droop occurs. “We hope that this will allow us to gain a general understanding of the droop so that we can understand and predict trends with respect to temperature, bandgap, and the asymmetry of electron and hole properties.”

D. Meyaard et al. Appl. Phys. Lett. 99 251115 (2011)