The role of leaked carriers in reduced LED efficiency

Light-emitting diodes use far less energy than traditional incandenscent light bulbs. However, at the high currents required for household lighting applications, LEDs suffer a significant loss in effiency known as efficiency droop. This droop is most noticeable in InGaN (indium gallium nitride) LEDs, which produce the blue and green light necessary for the white hues to which customers are accustomed. In order to overcome this obstacle in bringing InGaN LEDs to market, a concerted effort has been made over the last several years to understand the origin of efficiency droop. Although several mechanisms^1–5 have been proposed—including electron leakage and piezoelectric fields in quantum wells (QWs)—none are satisfactory. These models, while plausible, lack direct experimental evidence. The supporting data is indirect in most cases, and the models often rely on simulations that depend on unconvincing assumptions.

One promising candidate is carrier leakage: the escape of charge-carrying electrons from light-producing QWs. Carrier leakage in InGaN devices is a common problem because polarization in the multiple quantum well (MQW) region can reduce the barrier height and allow electrons to escape. The increase of carrier leakage with higher injection current makes this a leading contender to explain efficiency droop.

To determine the role, if any, of carrier leakage, we set out to make a quantitative determination of the leakage and its effect on droop.⁶ To do this, we illuminated an InGaN LED, operating in forward current mode, with a resonant excitation laser. By comparing the number of escaped carriers to the total number generated by the laser—the leakage ratio—over a range of forward bias (and injection current) values, we determined the amount of leakage as a function of injection current. We used two samples that exhibit markedly different droop behavior to determine the relationship between carrier leakage and efficiency droop.

Figure 1 shows a schematic of the experimental setup used in this study. When the electroluminescence is on, an excitation laser with 405nm wavelength illuminates the LED surface at a normal angle. The laser intensity is modulated by a mechanical chopper connected to a lock-in amplifier system. Among the carriers generated by the additional laser excitation, only those that escape from the quantum well can contribute to the change in the forward current. To quantify the number of laser-generated carriers, we calculated the absorption (A) of the laser in the LED as a function of applied voltage. The laser absorption was determined from the reflectance (R) and transmittance (T): A=1−(R+T).

Figure 1. Experiment schematic. A laser excites carriers in the LED. The number of these carriers is calculated by measuring the laser transmittance and reflectance. Leaked carriers are detected as an increase in the forward current (measured by the voltage drop across the resistor, R_L).

We measured a leakage current of 2–8% of the total generated carriers, which increases slowly as the injection current increases (see Figure 2). There are two ways to interpret these results. On the one hand, considering that the efficiency at 100mA (56A/cm²) droops, in general, by 20–40%, the observed 8% carrier leakage can be a significant cause of the efficiency droop. Furthermore, if the possible absorption of the laser in the p-GaN layer is taken into account, our resultant leakage ratio would become even larger. However, even at 100mA, the measured ratio does not dominantly account for the efficiency droop. In other words, while the carrier leakage does occur and can be significant, it is not enough by itself to explain the lowered efficiency. We confirmed this by comparing the leakage ratio in two samples that have clearly different efficiency droops. We found that the measured overall leakage is larger in the sample with a smaller droop. This further demonstrates that carrier leakage is not the only droop mechanism.

Figure 2. The ratio of leaked to total number of carriers increases along with injection current.

As part of ongoing efforts to improve the efficiency of InGaN LEDs at currents required for household lighting, we investigated the role that carrier leakage plays in the poorly understood efficiency droop. By illuminating an LED with an excitation laser and measuring the change in the forward current over a range of bias voltages, we were able to quantify the relationship between carrier leakage and injection current for two devices with different droop behavior. While we find that higher injection currents do lead to increased leakage rates, the number of leaked carriers cannot solely account for the lost efficiency. Our next step is to address the low temperature (<100K) behavior of the efficiency droop that some models invoke to explain the droop origin. By modifying our setup to include low-temperature measurement capabilities, we will investigate the temperature dependency of the carrier leakage.