As LEDs plunge to cryogenic temperatures, droop kicks in at lower drive currents due to diminished utilization of the active region.

A team of Korean researchers is proposing an alternative explanation for variations in LED droop with temperature. The researchers account for droop, the decline in device efficiency as the current is cranked up, with a theory involving carrier overflow, reduced effective active volume and saturation of the radiative recombination rate.

The researchers from Hanyang University, Korea, started by measuring the internal quantum efficiency (IQE) of blue and green InGaN LEDs at drive currents ranging from less than 1 µA to just below 10 mA at temperatures from 50K to 300K.

At cryogenic temperatures LEDs suffered from severe droop, attributed to carrier overflow. It is claimed that the degree of overflow is governed by the extent of the reduction in effective active volume and the subsequent saturation of the radiative recombination rate.

Measurements were made on blue and green LEDs with indium compositions of 20 percent and 35 percent and chip sizes of 350 µm by 430 µm and 400 µm by 400 µm, respectively. These devices were placed in a closed-cycle cryostat and temperatures measured at an aluminium plate that mounts to the TO package of the LED.

“The thermal resistance from the aluminium plate to the LED chip is not considered high, and even if there is a temperature gradient from the aluminium plate to the active layer of the LED, it shouldn’t be high for the given current level that the efficiency droop starts to occur,” says corresponding author Jong-In Shim from Hanynag University.

Shim points out that for temperatures below 100 K, droop kicks in below 0.1 mA, so power dissipating at the chip will be very small. He believes, therefore, that any differences between the actual and recorded temperatures for the LED will not be big enough to have impact on the conclusions that have been drawn from these measurements.

One of the findings of these measurements was that when the reduction in the temperature of blue and green LEDs was small, IQE increased and there was no significant change in the onset of droop. Shim and his co-workers believe that this occurs because decreases in temperature lead to a monotonic reduction in the number of defects in the quantum well that undergo non-radiative recombination.

When temperature falls further – below 200 K for blue LEDs and less than 250 K for their green siblings – device behaviour changes dramatically, with droop kicking in at a far lower drive current.

This led the team to draw two conclusions: The dominant droop mechanism undergoes a qualitative change at low temperatures; and Auger recombination is not a major cause of droop, because it cannot account for severe droop at low temperatures.

To shed more light on the cause of droop, the team measured electroluminescence spectra at various temperatures. In both LEDs, reductions in temperature led to the emergence of an emission peak around 400 nm (see Figure). This peak is claimed to originate from the overflow of electrons to the p-GaN cladding layer, where they undergo a recombination process involving a magnesium acceptor level.

 



Cooling unveils electron overflow

Shim and his colleagues believe that electron overflow is the dominant cause of droop at low temperatures. And they argue that the degree of this overflow gets more severe as the temperature plummets because the radiative recombination rate saturates at lower currents, and the non-radiative recombination rate reduces.

Saturation of the radiative recombination rate is thought to occur at lower currents, due to a reduction in the effective active volume. This volume diminishes as temperature falls, due to a reduction in the utilization of the active region, which is associated with degradation in carrier transport. Specifically, the combination of inferior hole transport at lower temperatures and the activation of fewer holes leads to a reduction in utilization of the wells, trimming the effective active volume. The upshot is an earlier saturation of the radiative recombination rate, and a greater overflow of electrons into p-GaN.

Measurements by the researchers also suggest that electron overflow is more severe in green LEDs than blue ones. One reason for this is that the green variants have greater indium, leading to wells with greater inhomogeneity and more indium clustering.

Competing theories

Shim and his co-workers are not alone in studying the influence of temperature on LED efficiency and using these measurements to draw conclusions on the origins of droop. For example, Fred Schubert’s group – which has measured the efficiency of blue LEDs at various currents and temperatures – claims that droop is caused by transport issues that stem from asymmetry in electron and hole concentrations, plus differences in carrier injection (see Appl. Phys. Lett. 99 25115 (2011)).

“It is true that transport is one of the factors that affects efficiency droop, but we think that there is another factor, which is more important, [saturation of the re-combination rate,” says Shim.

According to him, the model proposed by Schubert’s group is predominantly based on a p-n homojunction, and fails to consider the active region where recombination occurs. In particular, it fails to explain the following three experimental observations: Photoluminescence induces efficiency droop, LEDs of different colours are subjected to different levels of droop, and variations in device architecture impact LED behaviour.

Regarding his first point, Shim cites independent work by Osram and Philips Lumileds showing that resonant photoluminescence (PL) induces an efficiency droop similar to electroluminescence (EL). “This indicates that there is another factor in the efficiency droop, which is inherent in quantum wells,” says Shim.

Expanding on his second point, Shim adds that the ultraviolet LEDs incorporating p-type AlGaN are expected to show more carrier asymmetry than the p-GaN used in blue and green LEDs. “However, UV LEDs show less efficiency droop than blue green LEDs.”

Schubert can counter many of the issues raised by Shim. He argues that measurements performed by his team show that resonant-PL droop occurs at much higher excitation densities than EL droop. “Therefore, the resonant-PL droop does not need to have the same physical origin as the EL droop. I would not rule out carrier leakage as an explanation for the PL droop.”

According to Schubert, two factors can explain why green LEDs are more prone to droop than their blue cousins: Stronger polarization fields that supress carrier capture; and inferior p-type material quality, including lower p-type concentration and mobility that ultimately stems from the need for lower growth temperatures.

Schubert’s team have just published a paper with an analytical model for LED droop (see Appl. Phys. Lett. 100 161106 (2012)). With this model, a widening of the active region reduces droop – in other words, differences in the design of the active region do impact device behaviour. Claims based indium clustering that appear in the Korean conjecture are also controversial – Colin Humphreys’ group at Cambridge University, which has investigated GaN-based structures using scanning electron microscopes and atom probe tomography, claims that there is no indium clustering in quantum wells.

Shim’s response to this body of work is that indium can be locally segregated around point or line defects. He cites reports by others of clustering from atom probe tomography and near-field scanning optical microscopy measurements (see Stat. Sol. RRL 3 100 (2009) and Appl. Phys Lett. 87 161104 (2005), respectively).

He and his co-workers will continue to study LED droop. In particular, they plan

to carry out measurements of radiative and non-radiative recombination lifetimes, study the correlation between the degree of saturation in radiative recombination rate and efficiency droop, and investigate structures that suppress efficiency droop.

D.-S. Shin et al. Appl. Phys. Lett. 100 153506 (2012)