Reductions in external quantum efficiency at high current densities diminish when p-doped InGaN is inserted after the electron-blocking layer.

A Taiwanese collaboration has unveiled a novel LED architecture for combatting droop, the decline in device efficiency as current is cranked up.

The team’s modification to the standard LED structure is the insertion of a p-type InGaN layer after the p-AlGaN electron-blocking layer. Thanks to this refinement, the drive current associated with optimum external quantum efficiency (EQE) is increased by an order of magnitude to 316 mA. Meanwhile droop - evaluated as the percentage fall from the peak value of EQE to its value at 1A – plummets from 42 percent for a conventional control sample to just 7 percent.

These improvements in high-current LED performance are thought to result from mitigating the asymmetric carrier distribution in the device, thanks to enhanced hole injection and suppression of electron overflow.

When a conventional electron-blocking layer is used in an LED, the results are not ideal, because this barrier to electrons impedes the flow of holes into the active region. Making matters worse, at high current densities a significant proportion of electrons overflow into the p-type region, where they recombine with holes. This may be a non-radiative process, but even if it is radiative, it will be at a significantly different emission wavelength from that produced by the active region.

One approach to combatting droop, which has been shown to work by researchers at Philips Lumileds, is to use a very thick InGaN quantum well in the active region. The Taiwanese team has also built and studied these so-called double heterostructure LEDs. According to Sheng-Fu Yu from National Cheng Kung University, their performance is not particularly good, due to poor carrier transport and the strong quantum-confined Stark effect.

The team believes that it has uncovered a superior structure for addressing droop via a study that compares a conventional LED with those featuring additional undoped and p-doped InGaN layers inserted after the AlGaN electron-blocking layer. All the structures, which were grown with a Taiyo Nippon Sanso SR2000 MOCVD reactor on sapphire, featured a strain relaxation region comprising two pairs of 3 nm-thick InGaN and 12 nm-thick GaN

.

Although the LED with the p-doped InGaN layer before the AlGaN electron-blocking layer produces an inferior external quantum efficiency at lower currents, it is the most efficient at high drive currents

This strain relaxation section preceded the multiple quantum well – four pairs of 3 nm-thick In0.16Ga0.84N sandwiched between 12 nm-thick GaN barriers, followed by another 3 nm-thick In0.16Ga0.84N well and a 3 nm-thick GaN barrier. The control had an additional 5 nm-thick layer of GaN, while in the other two devices replaced the final GaN layer with 5 nm of In0.07Ga0.93N, either undoped or p-type with a magnesium concentration of 5 x 1019 cm-3. Onto all of these structures the team added a 20 nm-thick, p-type Al0.2Ga0.8N layer with a doping level of 1 x 1020 cm-3 and a 100 nm-thick p-type GaN layer with a doping level of 5 x 1019 cm-3.

Simulations with SiLENSe software made and sold by the STR Group suggest that InGaN insertion benefits the distributions of electrons and holes. With the standard LED design, at a drive current of 1 A for a 1 mm x 1mm chip, the leakage current in the p-side is 7 x 108 cm-3 and the hole concentration in the well nearest to the p-side is 9 x 1017 cm-3. In comparison, inserting undoped and doped InGaN before the electron-blocking layer slashes electron leakage in the p-side to 1.5 x 106 cm-3 and 9 x 105 cm-3, respectively, and increases hole concentration in the wells to 1.1 x 1019 cm-3 and 2.5 x 1019 cm-3, respectively. Real measurements prove that the addition of an InGaN layer, particularly if it is p-doped, is highly beneficial (see Figure). The conventional LED and that with an undoped InGaN layer produce a peak EQE of more than 40 percent at a drive current below 100 mA.

In comparison, although the device with the p-InGaN layer produces a significantly lower EQE in this regime, it overtakes the other two devices in the efficiency stakes at a few hundred milliamps and at a 1A drive current its output power can hit 950 mW, compared to about 800 mW and 700 mW for LEDs with an undoped InGaN layer and the control design, respectively.

One interesting feature of the LED with the p-doped InGaN layer is its low forward voltage. “This is a high-power chip, so I think its forward voltage should be judged at 350 mA,” says Yu. On that basis, the forward voltage is just 3.32 V, 0.3 V lower than the conventional design.

The team suspects that the decrease in operating voltage stems from a change in the location of the p-n junction that makes it more favourable for holes to be transported into the active layer. However, these engineers admit that they will have to carry out more experiments to confirm this hypothesis.

Efforts to date have been directed at understanding the relatively low EQE of the best device at lower current densities. Three weaknesses have been uncovered: Reversed current-voltage plots have unveiled a hike in the current of this type of LED compared to the others, indicating far greater defect-assisted tunnelling; analysis of current at low positive voltages shows severe tunnelling leakage, implying insufficient radiative recombination; and secondary ion mass spectrometry has revealed that the magnesium dopant atoms in the p-type InGaN layer have diffused into the active region.

Yu says that they will now try to improve the growth process to prevent magnesium atoms from diffusing into the active region.

R –M. Lin et al. Appl. Phys. Lett. 101 081120 (2012)